1

CHARACTERIZATION AND DYNAMICS OF AIR POLLUTANTS

IN THE SOUTHEASTERN MEXICO-US BORDER AREA

PROJECT NUMBER: A97-2

HENK L.C. MEUZELAAR1, GERARDO MANUEL MEJIA-VELAZQUEZ2,ROSAMARIA LOPEZ-FRANCO2

1CENTER FOR MICRO ANALYSIS AND REACTION CHEMISTRY, UNIVERSITY

OF UTAH, SALT LAKE CITY, UTAH, USA; 2CENTER FOR ENVIRONMENTAL

QUALITY, ITESM, MONTERREY, N. L., MEXICO

INTRODUCTION

Air quality throughout the Southeastern U.S.-Mexico Border Area, better known as the Lower RioGrande Valley (LRGV), is being threatened by rapid urbanization, extensive industrial and agriculturaldevelopment, and fast increases in border crossing vehicular traffic[1]. Yet, until recently, relatively fewair quality studies of the LRGV area had been published and emission inventory data for the LRGVarea were far from complete. In 1997 results of the multi-media Lower Rio Grande ValleyEnvironmental Scoping Study (LRGVESS)[2] started to provide systematic data on air pollutionsources, transboundary transport mechanisms and exposure risks. Also, preliminary emission inventorydata reported by Mejia and Rodriguez[3] enabled the ITESM team to perform a first assessment ofphotochemical pollution mechanisms[4].

In 1995 the University of Utah and ITESM started a collaborative effort, sponsored by the South WestCenter for Environmental Research and Policy (SCERP) and aimed at physical, chemical and biologicalcharacterization of fine particulate matter (PM10) in the LRGV[4]. In December 1995, 48 to 72 hr longscoping studies were carried out at 4 selected sites (Hidalgo International Bridge; Santa Ana WildlifeRefuge; Brownsville International Bridge; Matamoros Industrial Park). Typically, PM10 levels and sizedistributions were continuously measured around the clock. Particulate samples <PM10 werecollected on quartz fiber filters at 2-hourly intervals for subsequent laboratory analysis by means ofspecialized GC/MS techniques. Simultaneously, PM10 samples were taken for microbiologicalanalysis, meteorological parameters were recorded and a limited number of VOC samples werecollected and analyzed on-site using the University of Utah's mobile analytical laboratory withfield-portable GC/MS equipment.

2

Although detailed multivariate analysis of the extensive data sets obtained continued well into 1998 (dueto the limited availability of funds) preliminary evaluation of the voluminous data revealed modest overallPM10 and VOC type air pollutant levels (compared with earlier field tests in Nogales, Arizona)[5] withthe exception of one severe nocturnal PM10 episode at the Hidalgo site and unexpectedly high levelsof airborne fecal bacteria at the Brownsville site. The application of a diagnostic meteorological modelfor the LRGV provided, for the Hidalgo site, plausible explanations for the observed severe PM10episode which apparently followed the passage of a cold front accompanied by low inversion layerheights and slow drifting of urban dust trapped above Reynosa around the Hidalgo international bridgearea. The high concentrations of airborne fecal bacteria at the Brownsville international bridge arethought to originate from the Rio Grande River, although the specific aerosolization mechanism is asyet unknown.

Based on the preliminary VOC and PM10 findings of our scoping studies, we decided to perform afollow-up field study focusing on detailed physical (including size distribution) and organic chemicalcharacterization of PM10 at selected receptor sites on both sides of the border. In addition, weattempted to simultaneously model and monitor criteria pollutants such as NOX, O3 and SO2 in orderto understand the origin and dynamics of both primary and secondary PM10 in the LRGV section ofthe US/Mexican border. Consequently, in March, 1998 a second, 5-day field study was undertakenon both sides of the US/Mexican border between the twin cities of Brownsville/Matamoros andMcAllen/Reynosa. During this study, measurements and samples were taken in Reynosa, Rio Bravoand Matamoros, Mexico, as well as in Hidalgo and Brownsville and along the freeway betweenBrownsville and McAllen in Texas. A diagnostic meteorological model was applied to the region tosimulate wind patterns during the sampling period.

The work reported here consists of two parts: (1) development of a comprehensive criteria pollutantmodel for selected LRGV areas involving the integrated use of emission, dispersion and photochemicalsubmodels and with attempted validation by field monitoring data obtained in March 1998; and (2)development and testing of a novel time-resolved PM10/PM2.5 organic chemical characterizationapproach involving a combination of fast, sensitive physical and chemical receptor monitoringtechniques as well as the use of principal component analysis techniques to detect and identify thedominant emission sources for the selected sites and time windows, with a special focus on the Hidalgointernational bridge data obtained in December 1995.

3

OBJECTIVES OF THE RESEARCH

To identify the chemical nature and concentration of major organic air pollutants [gases, (semi) volatiles,particulate matter] as well as their diurnal and seasonal variation and the effects of changes inmeteorological conditions.

To identify major biological components and correlate these with known seasonal variations as wellas the and the effects of changes in meteorological conditions.

To apply advanced dispersion models and modeling techniques to study the dynamics of differentclasses of air pollutants in the region, based on GIS data as well as on measured data obtained atrepresentative sites and dates.

To further develop the GIS initiated during the first year by adding population, fuel consumption, landuse, emission and source location and dynamics of air pollutants.

To make a preliminary assessment of the environmental impact of air pollutants in the region in directcollaboration with local and regional authorities on both sides of the border.

RESEARCH METHODOLOGY/APPROACHES

To accomplish our objectives, we planned a field monitoring trip to obtain samples and monitor airquality; we used a diagnostic meteorology model to estimate and analyze wind conditions in the area;and we used the CIT photochemical model to make a preliminary assessment of the dynamics of airpollutants in the region. Physical measurements included particle size distribution determinations witha six channel CLIMET aerosol counter and meteorological measurements (temperature, pressure,humidity, wind speed and direction) with a Davis model III weather station. Planned on-site chemicalanalyses involved NOx and O3 measurements, as well as VOC/SVOC speciation using a novel GC/MStechnique with a miniaturized, fast GC and Curie-point desorption modules. Several problems wereencountered during the 72 hr measurement period (March 11-14, 1998), namely: (1) frequent rainshowers (reducing pollutant levels and necessitating longer collection periods while turning monitoringsites into mud pools); (2) electrical failure of the ozone analyzer; and (3) intermittent leaks in theVOC/SVOC desorption inlet. As a result, the number of PM10 collection sites, originally anticipatedto be as high as 20, had to be reduced to only 6.

4

Fine Particulate Matter Characterization

On-site monitoring and sample collection of fine particulate matter and VOC’s were performed withthe University of Utah's mobile pick-up truck transportable laboratory. (Figure 1) This mobilelaboratory is designed to be a platform for a range of physical and chemical air pollutantcharacterization methods, including PM10 counting, collection and organic analysis of PM10adsorbates. In particular, by incorporating a field-portable GC/MS system (Figure 2) with a thermaldesorption/pyrolysis inlet, this laboratory allows the speed and sensitivity of GC/MS techniques to beemployed for time-resolved characterization of PM10 adsorbates and VOCs, in near real-time ifdesired. Instrument power was provided by a 400 Ah battery bank with 2000 W batterycharger/inverter, and a 4 kW propane-driven generator. A Peltier-cooled refrigerator box was usedfor sample storage.

PM10 was collected on QM-A type quartz fiber filters (QFFs) from Whatman, (Hillsboro, OR) orQFFs manufactured by Pallflex (2500 QAT-UP grade, Pallflex Products Co., Putman, CO).Collection of particles was achieved through the use of a modified dichot sampling inlet. In thissampler, aerosol is drawn through an isokinetic sampling inlet with subsequent impactor stage whichdeposits particles greater than 10 :m on a plate immediately in front of the impactor nozzle. Particlessmaller than 10 :m are carried vertically upward by the air flow (at 20 l/min) and down vent tubes tothe filter where they are collected (at 2-hourly intervals) as schematically depicted in Figure 3. Theimpactor is contained in a stainless steel delivery tube which channels particles #10 :m to both thefilter holder and to a Climet model CI 208C particle counter which counts particles #10 :m in eightseparate particle size ranges covering 0.3–10.0 :m diameter. Aerosol concentration and particle sizedistribution measurements were performed at 1–4 minute intervals with a specially designed computerinterface for continuous data collection. Simultaneous meteorological parameters of wind speed, winddirection, temperature and barometric pressure were measured with a Davis Weather Monitor II(Davis Instruments, Hayward, CA) interfaced to a PC workstation. Sampling ports were locatedapproximately 5 ft above the roof of the mobile. Examples of particle density and meteorological dataobtained are shown in Figure 4.

The collection and concentration of VOC was accomplished through the use of multisorbent adsorptiontubes. Quartz glass tubes of 2 mm diameter contained 10 to 20 mg of three carbon-based adsorbents;Carbopack B, Carobopack X, and Carboxen layered in 2:1:1 ratios, respectively. This combinationof sorbents allowed collection of a range of compounds from C3 to C12. A flattened helicalferromagnetic heating wire with a Curie-point temperature of 510° C in the center of the tube andsorbents provided the means for fast thermal desorption of adsorbed compounds (Figure 2b).

On-site chemical analyses involved VOC/SVOC speciation as well as PM10 adsorbate analysis bymeans of GC/MS. Examples of the type of GC/MS data obtainable are shown in Figures 5a (VOCs)and 5b (PM10 adsorbates). More detailed off-site chemical analyses of quartz fiber filters were

5

performed at the University of Utah Center for Micro Analysis with a standard HP GC/MSD equippedwith a special Curie-point desorption/pyrolysis inlet. Detailed descriptions of these techniques havebeen given elsewhere[4,5,6] and are outside the scope of the present report. Multivariate analysis ofcombined physical and chemical measurement data involved the use of principal component analysis(PCA) techniques in combination with graphical rotation methods[5].

Microbiological Characterization

Biological components were collected from a sample port above the mobile laboratory using a GellmanScience (# cat. 66191) membrane filter with of 0.45 mm and 47 mm diameter pores. Filters wereconnected to a vacuum pump (900-13-58C) operating at 40 l/min. The membranes were placed ina sterile filter holder and exposed for two hrs each. After exposure, the membranes were placed inanother sterile container and kept under refrigeration until analyzed at the Microbiology Lab at ITESM.Exposed membranes were placed in sterile Petri dishes with the appropriate medium: for total coliformsMF-Endo Broth (Gelmans Science # 68201) was used; and for fecal coliforms M-FC broth withrosolic acid (Gellmans Science # 68101)was used. A 24 hour period of incubation followed, at 37 Cfor total coliforms (Amerex Gyromax 703), and for fecal coliforms at 45 C (Shelab mod. 1535). Theplates were removed from the incubation and read for the bacterial analysis and count. For othermicrobial growth, samples were incubated an additional 24 h and plates analyzed for presence of UFC(colony formation units). Prior to incubation, filters were carefully cut in half so we could proceed toanalyze for the total and fecal coliforms. After the bacterial analysis, the plates were incubated 24 a48 h longer to induce other microbial (fungal) growth. Microscopic observations were made using aOlympus Vanox foto and video microscope with equipped with a Hamamatsu Argus-20 imageanalyzer. The microbial growth was recorded using Polaroid 667 film.

Meteorological Modeling

Cross border transportation of air pollutants is by wind. Meteorological models are an important toolto estimate wind magnitudes and directions in a given region using data obtained from meteorologicalstations. Two types of meteorological models are commonly used for this purpose. One type of modelis a prognostic model, which is based on mass balance, energy and momentum equations. This typeof model can predict the evolutions of wind and meteorological conditions starting from conditionsknown at a given time. With these models the influence on meteorological conditions of changes in landuse can be evaluated. The other model type is a diagnostic model, which is based oninterpolation/extrapolation techniques. Diagnostic models only estimate the meteorology conditions inthe domain from data obtained at a given time. For the purpose of this study, a diagnosticmeteorological model was used to estimate wind patterns in the region. Data of US and Mexicanairports in the region as well as the data collected during the monitoring trip were used as input to themodel. The results of this model were used in the photochemical model to study the dynamics of NOX,SO2, and ozone.

6

Modeling of Photochemical Pollutants

Modeling techniques used by the ITESM group included the application of a diagnostic meteorologicalmodel, the use of a GIS to generate a land use and an emission data base, and the application of aphotochemical model. The GIS was used to create a database of meteorological, emission, andpredicted concentration. The data base allowed us to achieve two objectives: (1) to create input datafiles for the photochemical model; and (2) to display wind, emission, and concentration data in a mapof the LRGV for better understanding of the results.

The CIT Photochemical Model[7] was applied to study the dynamics of pollutants in the region. Inputdata files included emissions, wind fields, incoming solar and UV radiation and land use. Outputs ofthe model are maps of pollutant concentrations in the region and time series for different pollutants. Inthis report the predicted SO2 and O3 concentrations in the period of March 12 to 14, 1998, arediscussed. Predicted NOX concentrations are compared with data collected during the monitoringstudy on March, 1998.

Emissions in Mexico were estimated using the Mobile5 Juarez model developed for Cd. Juarez byEPA[8, 9]. Other emissions in the area were estimated in a previous study based upon fuel consumptionand emission factors[3]. Stationary source emissions in Texas were obtained from TNRCC. Mobilesource emissions were not available and were estimated using a regression analysis of population andemissions of other counties in Texas made in a previous study of the border area[10]. Wind fields werereconstructed with a diagnostic meteorology model developed to interact with the CIT model. Dataobtained with the monitoring station and from the airports were used for this purpose. Land use dataof the LRGV were obtained from INEGI and from the USGS in digitized form. Solar radiationmeasurements were not available and, therefore, data were estimated from the geographicalcoordinates and calculated incoming solar radiation [11].

PROBLEMS /ISSUES ENCOUNTERED

Several problems and issues were encountered in carrying out the objectives of this research. For thedevelopment of the study, especially in completing the GIS and for model development, emission andmeteorological information is needed. However, little of this information is available for the U.S. andonly meteorological data exists for Mexico. The emission source information in the U.S. is reportedfrom point source emissions only. Mobile source and area source emission data is not available. Thesetypes of emission data were estimated for our test location based upon population and point sourceemissions reported from some Texas counties. Emissions for Mexico had to be estimated from fuelconsumption data and vehicle fleet numbers in the test area.

7

In addition, there is a severe lack of meteorological data in the LRGV. This makes it difficult to studyand determine pollutant pathways. For this study, we used meteorological information obtained fromthree airports located in the region in addition to the data we collected in the field. Although our resultsseem reasonable, if more monitoring sites existed in the LRGV, the increased data input would certainlyimprove our accuracy.

Chemical data collected were limited due to several factors. During the monitoring field trip the ozoneanalyzer failed because of an electrical problem, resulting in the lack of ozone concentrationinformation. Leaks developed in the SOV/VOC inlet. And severe, frequent rain showers (which areextremely unusual in their duration and intensity for that time period) decreased particulate, bacterialand air pollution concentrations. Therefore, results are surely lower than, and not necessarilyrepresentative of, normal or average values in the LRGV during that time. As a result of theseproblems, the number of monitoring sites, originally anticipated to be as high as 20, had to be reducedto 6 PM10 and 7 NOx measurement sites (see Figure 6).

RESEARCH FINDINGS

Land Use and Monitoring Sites

Land use in the LRGV is shown in Figure . This figure was made with ARC/VIEW and shows thegeographical location of the most important cities in the area. McAllen and Brownsville are located onthe U.S. side and Reynosa and Matamoros on the Mexican side of the international border. The landuse displayed in Figure 3 shows the different classifications used in the U.S. and Mexico. The data wasobtained from data bases of the USGS and INEGI. An important problem found creating this figurewas that the scale used on each country is different, then integrating the data with GIS was not a simpletask. Although most of the land is used for agriculture, different classifications are used in each countryand, for air quality modeling purposes, a unique classification should be used. This is important sincedeposition velocities for different pollutants are estimated depending on land use. The figure also showsthe grid used in the diagnostic meteorological and the CIT photochemical model. Using GIS software,the land use data was assigned to each grid cell and the result of this process was a file to be used inthe photochemical model.

The monitoring sites of the field trip are shown in Figure 6. Sites A and C correspond to locations inReynosa and the Hidalgo Bridge, Site D was in Rio Bravo, where a Power Plant is located, and sitesB and E correspond to Matamoros and the Brownsville Bridge. Numbers 1 to 8 show the sitesmonitored along the highway from Brownsville to McAllen.

8

Wind Patterns in the LRGV

Transport of air pollutants in the LRGV is dominated by air flowing in from the Gulf of Mexico, fromeast to west, although at night quiet periods are common and sometimes the wind blows from the landto the sea ("land breeze"). During winter time, cold fronts coming from the north transport pollutantsfrom west to east at ground level, while the warm air from the Gulf of Mexico flows in the oppositedirection in the upper part of the atmosphere. This effect may cause periods with high concentrationsof particles and other air pollutants in some areas of the LRGV, especially when occurring incombination with low mixing heights.

Meteorology data from the Matamoros and Reynosa airports, and from US monitoring stations in theLRGV were used in the meteorology model to estimate wind field vectors in the region. The dataanalyzed showed dominant winds coming from the Gulf of Mexico, from east and the southeast, asshown in Figures 4, 5, and 6, for March, 12, 13 and 14, 1998, respectively. The results of thediagnostic model are obtained for each hour. In the figures the wind fields at 3:00, 9:00, 15:00, and21:00 hours are displayed. All figures show predominant wind from the East at 9:00 hrs and from theSouth East at the other hours, except at 3:00 hrs, March 13, that shows wind form the West in thewestern side of the domain, in the region of Reynosa/Hidalgo. A possible explanation is that during themonitoring trip we had rain and periods of calm, then small variations in wind, or even mistakes inreported data, may reflect wrongly estimates in wind patterns when a diagnostic meteorology modelis applied. Also, it is not simple to obtain good estimates of wind patterns in the region when fewmeteorology stations exist in a domain of study. Nevertheless, the results obtained are consistent withthe general wind patterns registered in the area.

PM10 Physical Size Distribution

Table 1 shows the average particle size distribution for the different sites and monitoring periodscovered during the field trip on March, 1998. The data collected with five of the six channels of theCLIMET are shown in the first column, and covered the range from 0.3 to 10 microns. Three of theseranges covered PM2.5. The sixth channel covered the number of particles larger than 10 microns andis not shown in this table. The average number of particles per 2.0 ft3 sampled are shown in the secondcolumn, N(dp). The flow rate of the CLIMET was 1.0 ft3/min. The volume of particles in each range,Vi, was calculated assuming the particles were spherical[11]. The total volume, Vt, is the sum of the Vi,hence, Vi/Vt represents the volume fraction of each range of particles. Figure 7 graphicallysummarizes and compares average particle size concentration, divided into fine, coarse, and TotalPM10, along with meteorological information from the field trip to Brownsville, Texas in Dec. 1995. Further details about this figure may be found in the attached pre-print.

In general, small particles are large in number but big particles are more important in volume and,consequently, in mass. On the other hand, it is well known that PM2.5 represents the fraction of PM10

9

most hazardous to health. During the monitoring trip we found that PM2.5 represents approximately16 to 20% of PM10 in Hidalgo and Brownsville, Texas. Similar values were found in Matamoros,Mexico. In the case of Reynosa and Rio Bravo we found that PM2.5 accounts for 55 and 41.3% ofPM10 respectively. These values were found during monitoring periods that corresponded to hightraffic near the sites: 6:00 to 12:00 hrs in Reynosa and 16:00 to 18:30 hrs in Rio Bravo. These twocities are located downwind, have many unpaved streets and receive most of the pollutants emittedfrom the highway between Matamoros and Reynosa. In Reynosa, during a monitoring period from 0to 2:00 hrs it was found that PM2.5 accounted for 16.3% of PM10. This value was obtained at nightafter a rainy period and in a monitoring site with very low traffic. From these results, we observe thatPM2.5 could be an important fraction of PM10 in locations downwind of the LRGV. However, toobtain more reliable results and conclusions more data collected during different seasons of the yearare necessary, as well as mass concentration of PM10 to obtain concentrations of PM2.5.

PM10 Biological Composition

The sampling sites, membrane exposure and precise location are shown in Table 2. The number oftotal colonies (UFC) or colony formation unit are given in the Table 3. The biological composition ofPM10 accounts for the bacterial colonies. We found a relatively low biological load in the samples.That is, for each sample, at least 4,800 L of air were filtered and few colonies were recorded. Thebacterial composition for fecal coliforms were mainly bacilli of 1-2.5 X 0.5 micron size while for thetotal coliforms, a mixture of cocci ( 0.5 microns diam.) and bacilli were found. The presence of thefecal coliforms, although low, due perhaps to the presence of the rain during the sampling, reflects poorconfinement of organic waste from domestic or industrial origin.

For the fungi, it is very likely that the spores of these organisms were present in the samples. Penicillium and Aspergillus have the smallest spores of 10-15 microns, while pieces of vegetativemycelium or resistant chlamydospores are also common means of dispersion and colonization.

The health interest in these fungal species is that, although they are common components of the air flora,they often are the cause of several respiratory infections mainly in the immunocompromised population.

A careful microscopic analysis of the membranes was done to look for other biological components(pollen grains, worm's eggs, etc.) however we found nothing that could resemble other biologicalcomponents, perhaps due to the constant wet conditions that have help to precipitate and wash off theairborne particles.

10

Chemical Characterization Results

Figure 4 shows a total ion chromatogram of VOCs obtained in Reynosa from a multisorbentadsorption tube. Labeled peaks identify desorbed compounds from C3 to C12. The first peak is broaddue to the large amount of water in the air and also contains acetone, a common solvent encounteredin urban VOC samples. Many of the Aliphatic and aromatic hydrocarbon peaks in suchchromatograms must be attributed to fugitive gasoline and/or completely burned transportation fuels.

Particulate chemical analysis data collected from Brownsville are shown in Figure 4. Several classesof compounds including phenols (represented by m/z 153 and m/z 165), phthalates (represented bym/z 149) and polynuclear aromatics were detected. Although the rain cleaned many particulates fromthe air, the profile and compounds found in this 2 hour filter sample are qualitatively similar toparticulate samples we have collected in different locations and at different times.

By and large, however, the VOC and PM10 adsorbated data obtained during these very rainy dayswere too low in intensity and too fragmented with regard to sampling locality and sampling time to allowsystematic analysis by means of multivariate techniques. Consequently, we decided to further analyzeGC/MS, particle density and meteorology data during earlier visits to the same LRGV region in orderto obtain a more complete overview of the major sources of air pollutants, their primary circadianrhythms and the main meteorological parameters affecting air quality, particularly in winter (earlyDecember).

Extensive chemical identification was completed on particulate matter GC/MS data obtained fromscoping studies in Nogales (1991), Calexico (1992), and Brownsville and Hidalgo (1995). In addition,multivariate analysis of the 1995 scoping study data for the Brownsville and Hidalgo international bridgesites has been completed and some results for the latter site, also included in our 1998 field study arediscussed. A more thorough discussion and explanation can be found in the attached SCERPmonograph chapter.

Multivariate Data Analysis Results

Table 4 shows the variables included in the final principal component analysis of the Hidalgo PM10scoping study involving 24 samples obtained at 2-hourly intervals. Note that this includes organicchemical compounds and meteorological parameters as well as PM10 and PM2.5 density estimates.After Varimax rotation, only 5 principal components were needed to explain nearly 80% of the totalvariance in the data set. The loadings for these five (Varimax-rotated) factors are listed in Table 4 andreveal a relatively well-behaved clustering of the variables along the different principal component axes.Fortunately, a reasonable chemical and physical interpretation of the first four factors appears to berelatively straightforward and is in good agreement with the dominant trends observed by Mukerjee

11

et al. in their Varimax-rotated principal component analysis of inorganic PM10 characterization datafor the LRGV region[12]. Figure 11 shows the power of the time-resolved PM10 analysis approachin that it provides the opportunity to help tie observed receptor sample patterns to possible emissionsources on the basis of known circadian human activity cycles and events (e.g. traffic peaks, wasteburning) or observed meteorological patterns and events.

It should be noted that the factor scores in the abscissa of Figure 11 are standardized, therebypermitting a rough estimate of the statistical significance of the various events. Assuming multivariatenormal distributions, both the (inferred) "urban dust" event in Figure 11a which followed the arrival ofthe cold front on day 2, and the "biomass burning" (primarily hardwood markers dominated) event inFigure 11c can be classified as being well in the 4 sigma range. The high amplitude of both events maycreate the mistaken impression that the corresponding chemical markers were only detected duringthese events. In fact even when leaving out these extreme episodes, the nature of the underlyingparameter associations does not change much at all. Further note that the 1st principal component("automotive emissions"; Figure 11b) shows regular, traffic-peak-related fluctuations dominated by theafternoon rush hours when most of the traffic was passing on the same (downwind) side of the bridgethat the mobile lab was stationed. Although the proposed "urban dust" factor is speculative at this pointsince the peaks at m/z 239 (abietic acid - prominent filler in car tires) and at m/z 306 (tetraphenylene- oily road stabilizer?) have not yet been positively identified in local source profiles, and the "wasteburning" events are inferred from the prominent contributions of alkylphthalate type plasticizers , the"automotive emissions" and "biomass burning" events are firmly rooted in the pioneering GC/MS studiesof well defined source samples described by Rogge et al[13]. Also, it should be mentioned that theinferred, broad "waste burning" event in Figure 11d was primarily characterized by the presence ofseveral different types of plasticizers plus a fire retardant. Finally, the authors would like to point outthat very similar "urban dust", "automotive emissions", "biomass burning" and "waste burning" factorloading and score behavior was observed at the Brownsville international bridge site data obtainedduring the same 1995 scoping study as well as in two earlier December time window studies at theUS/Mexican border (Nogales, Arizona 1991 and Mexicali, California 1993), thereby suggesting amarked degree of similarity in major PM10 sources along the border and further supporting thechemical and physical significance of the numerically extracted principal component patterns.

Sulfur Dioxide, Ozone and NOX Dynamics

Emission sources in the LRGV include mobile sources in Mexico and US, small stationary sources inthe Cameron and Hidalgo counties in the US, a PEMEX Refinery in Reynosa and a power plant in RioBravo, in Mexico. Several uncertainties exist in the estimated emissions used to create the data files forthe CIT photochemical model. These uncertainties include aspects like: lack of data to estimate mobilesource emissions from both the US and Mexico, an official emission inventory of stationary sources inMexico is not available yet, biogenic emissions from vegetation are not known and were estimated formland use, distribution of mobile emissions in highways is not well known, emissions from diesel trucks

12

were estimated, and diurnal variations in emissions of mobile sources was not considered. All theseuncertainties represent research areas to improve the data necessary to obtain more reliable results.Nevertheless, the results found in this work represent general trends of transport of air pollutants andpotential areas affected by high levels of pollution.

The region under study, the domain, and computational mesh covered with the photochemical modelcovered a surface of 180 km by 180 km, with cells of 5 km by 5 km, i.e., 1296 cells. Five layers weredefined in the vertical dimension which makes a total of 6480 cells. A GIS data base was loaded withthe necessary emission and land use data files create the input files to run the CIT PhotochemicalModel. The files were created using a procedure developed within the GIS for automatically transferdata to each cell. This procedure was made in the computer, minimizing numerical error when data areassigned to each cell and processing time.

The results of the model give average hourly concentration data of the different pollutants in each cellin the three spatial dimensions. Understanding results given as lists of numbers is very difficult andusually visualization techniques are used to facilitate this procedure. In this report, the predicted SO2

and O3 concentrations in the different monitoring sites during March 12 to 14, 1998 are analyzed.Figure 12 shows plots of SO2 concentrations for 5 sites in the domain: Reynosa, Hidalgo bridge,Matamoros, Brownsville, and Rio Bravo. The data displayed in this figure show that the higherconcentrations of SO2 are in the area of Rio Bravo, Reynosa and Hidalgo. This is expected since airis blowing from the sea and SO2 emissions of the Emilio Portes-Gil power plant and the refinery ofPEMEX in Reynosa are transported West of the source. In Matamoros the SO2 concentration inMarch 12 is high and decreases in March 13 and 14. This may be caused by an initial overestimationof SO2 emissions in the city. It is important to note that the higher concentrations predicted by themodel are close to 50 ppb which is well below the Mexican air quality standard for SO2 of 130 ppb.Furthermore, if sulfur content in fuel is lower than the values assumed we may expect lower SO2

concentrations in the area. Fluctuations in SO2 concentrations are mostly caused by change inatmospheric stability during the day and the night. During the day dispersion increases in theatmosphere and SO2 concentration increases at ground level in Hidalgo, Reynosa and Rio Bravo. Atnight the atmosphere becomes stable and, in the same locations, the ground level concentrationsdecrease.

Emissions of mobile and industry sources have a great direct impact in CO, HC and NOX

concentrations, which indirectly react to produce O3. Concentrations of this pollutant for the same 5sites discussed in Figure 12 are shown in Figure 13. This figure shows the same levels of predictedozone concentrations in nearby sites. This is explained since mobile source pollutants are emittedmainly in the cities and highways along the domain. We can appreciate higher concentrations of O3 insites located downwind - Hidalgo, Reynosa, and Rio Bravo - as in the case of SO2 concentrations. Inthe figure we observe that ozone concentration increases for March 12 to 14 in Hidalgo, Reynosa, andRio Bravo. This may be caused because the initial concentrations of ozone were below than those

13

estimated form the emissions with the photochemical model. Another important aspect is the uncertaintyin estimated emissions and their distribution. As expected, the figure shows the cycle in ozone formationduring the day. Ozone is formed by photolysis of NO2 and, consequently, its concentration increaseduring the morning, peaks early in the afternoon, and decreases at night. High levels of ozone may beencountered at night if it is not dispersed in the atmosphere or if it is not consumed by NO when theconcentration or emissions of this pollutant is low at night. We can observe that the model predictspeak concentrations of ozone around 180 ppb, which is above the Mexican and US air qualitystandards for ozone of 110 and 120 ppb respectively. Again, emission may be overestimated causinghigher ozone concentrations than expected. However, actual data in the domain are necessary in orderto validate the results of the photochemical model.

The results shown in Figures 12 and 13 are for locations within urban areas. The estimatedconcentrations in rural areas in some regions downwind of emissions and at some times of the day showto be higher than in urban areas. An important limitation is that these results obtained with the modelwere not validated for ozone since actual air pollutant concentration data do not exist in the region forthe period of study, except for the NOX data collected during the monitoring trip.

In the study of March, 1998, NOX concentrations were measured at the different sites. The results ofthe predicted values with the photochemical model and the actual concentrations measured are shownin Figure 14. In this figure it is observed that the model predicts concentrations of NOX in Brownsvilleand Matamoros that exceed the one hour air quality standard of 210 ppb, while measuredconcentrations of NOX and values predicted for other cities are below this standard. Reasonablecorrespondence was observed between predicted and measured NOX concentration profiles or trendsat some sites. However, predicted values exceed in general measured values. Interestingly, Table 5shows that the values measured in the highway from Brownsville to McAllen exceeded the valuespredicted by the model. This may be caused because the model predicts the average concentration ina grid cell, while measured values correspond to specific sites. In the case of the concentrationspresented in Table 3, they were measured close to highways and it is expected that high values of NOX

could be measured. The discrepancies between measured and modeled NOX may be due to the lackof complete emission inventories (particularly the fact that US emission data are summed per countyand not specified by highway), the daily variation and spatial resolution of the data, the highly unusualweather conditions, and the lack of sufficient data points to fully calibrate the models.

Photochemical models are important tools to study the dynamics of air pollutants in a given region. Inthis work we analyzed the dynamics of ozone in the LRGV. Figure 10 shows maps of ozoneconcentration in ppb at 3:00, 9:00, 15:00, and 21:00 hrs on March 13, 1998. The figure shows thatozone concentrations, as expected, are higher downwind of major NOX and HC emission sources. Infact, the model predicts that ozone concentration are higher in rural areas than in urban areas due tothe transport of ozone precursors (HC and NOX) downwind. Figure 10 also shows the diurnal variationof ozone concentrations in cities. Ozone increases during the day, but at night fresh NO emissions

14

depletes ozone in urban areas and areas surrounding highways, while in rural areas concentrationsdecrease due to dispersion. The three red spots shown at 21:00 hrs show the location of major urbanareas - Matamoros/Brownsville, Harlingen, and Reynosa/Hidalgo/McAllen. The model predicts ozoneconcentrations well above ozone national standards for 1 hr average are 120 ppb in the US and 110ppb. Although few field data points are available to compare results as was shown in Figure 6, we mayassume that the values predicted in the region may be high. There are several reasons that may explainthis behavior, some of them are that, although emissions were distributed spatially in each cell of thedomain according with their geographical location, the emissions were assumed to be constant duringthe day and in order to obtain reliable results daily variations must be considered for the differentsources. Also, emissions may be overestimated causing that the model calculates higher concentrationsof ozone. The HC have different reactivities to ozone formation and in this study we considered atypical speciation. Results are improved when speciation for the different sources are available to beconsidered in the model.

Despite the lack of emissions and meteorology data, the results of the diagnostic meteorology modeland the CIT photochemical model give a first order estimate of levels of pollution in the area and clearlyshow that pollution may be higher in rural than in urban areas.

CONCLUSIONS

C The results of this study show that the LRGV has in general low levels of PM10. Dependingon meteorological conditions these levels may increase showing high levels of concentrationduring some periods of the day. Differences may also be evident with seasonal variation.

C Time resolved analysis makes it possible to distinguish car emission sources from othercombustion sources such as wood burning or food preparation. It also can separate andelucidate source rhythms as diurnal circadian (car traffic, food preparation), irregular (woodburning) or incidental (urban dust).

C Size distribution results show that PM2.5 seems to be an important component of PM10. Itis important to obtain data for longer periods and for different seasons in the year to achievemore reliable results about the size composition of PM10.

C The multipoint sampling and analysis technique enables mapping of pollutant concentrationsthereby integrating receptor based and source monitoring models.

15

C Combining organic FPM analysis with meteorological measurements and particle concentrationusing PCA provides essential information regarding the origin and behavior of FPM type airpollutants.

C In spite of being very wet for the sampling season, we detected the presence of bacteria (cocciand bacilli) that falls in this category. Also, due to the wet season the amount of fungal coloniesthat grew in all the plates was quite large. This condition was probably increased due to thefavorable weather conditions for these kind of microorganisms.

C To obtain better calculated values of ozone, SO2 and other pollutants is necessary to havebetter emission estimates of stationary and mobile sources, as well as their diurnal and spatialvariation. Also, it is important to have information about the type of HC emitted by the differentsources or, at least, information of the type of fuel burned by each source in order to make anestimation of the HC emitted with models suggested by EPA.

C It is necessary to obtain actual concentrations of air pollutants to validate predictedconcentrations of air pollutants with the photochemical model.

C Usually air quality data are collected in urban areas. However, transport of pollutants to ruralareas may cause higher concentrations in these areas.

C Technical and logistic achievements were: trouble-free crossing (twice) of the US/Mexicanborder with a fully equipped mobile laboratory and successful (nearly around-the-clock)operation of a mobile laboratory on both sides of the US/Mexican border with a binationalteam of investigators while covering 6 sites over a 200 mile distance within a 72 hour period.

C It is convenient to define a common classification of land use for both countries as well as todefine similar scales in the maps produced in each country to facilitate data transference andstandardization.

RECOMMENDATIONS FOR FUTURE RESEARCH

C Perform comprehensive air pollutant modeling and measurement studies covering a broadrange of gaseous, volatile and particulate matter pollutants as well as meteorological data.

C Integrate research teams with Mexican and U.S. scientists to perform air quality studies in theborder area to obtain a more reliable approach to the problem, instead of workingindependently on both sides of the border.

16

C It is important to standardize data formats to be shared, transferred, and used by bothcountries.

C Collect samples and data during a dry season so seasonal variation may be studied

BENEFITS OF RESEARCH PROJECT

General benefits include:

The results of this project show the importance of developing of reliable emission inventories. Becauseof the fast development of the area air quality is likely to deteriorate unless preventative actions areconsidered by U.S. and Mexican authorities. The results of this project are provide useful newinformation to help understand the dynamics of air pollutants and to design environmental policies forthe area by local authorities.

Specific benefits include:

C One paper related to results of this project was accepted to be published in the proceedingsof A&WMA Annual Meeting.

C Results of the project were presented at the SCERP 1998 meeting and will be presented atthe 1999 A&WMA Annual Meeting.

C Three Mexican graduate students and one American graduate student were involved in theproject.

C A monograph describing some of the results of this project and previous SCERP projects isscheduled to be published in a volume of SCERP sponsored research on air pollution. (seeattached preprint)

ACKNOWLEDGMENTS

The research reported here was supported by the Southwest Center for Environmental Research andPolicy (SCERP). We thank the personnel of Secretaria de Desarrollo Social de Tamaulipas for their

17

valuable support during our monitoring trip. The information supplied by TNRCC, Subsecretaria deEcologia de Nuevo Leon and Secretaria de Comunicaciones y Transportes is gratefully acknowledged.Support of Universidad de Reynosa is also acknowledged.

REFERENCES

1. Gilbreath, J. Planning for the Border's Future: The Mexican-U.S. Integrated BorderEnvironmental Plan. U.S.-Mexican Occasional Paper No. 1, U.S.-Mexican Policy StudiesProgram, LBJ School of Public Affairs. The University of Texas at Austin, Austin TX, March,1992.

2. Special Issue. Environmental Aspects of the Lower Rio Grande Valley; Mukerjee, S. Ed.;Environment International, 1997, 23, 593-744.

3. Mejia, G. M.; Rodriguez, M.: "Characteristics and Estimated Air Pollutant Emissions of theIndustry and Vehicles in the Matamoros-Reynosa Border Region," Environment International,Vol. 23, No. 5, 1997, 733-744.

4. Mejia, G. M.; Meuzelaar, H. L. "Characterization and Dynamics of Air Pollutants in theSoutheaster Mexico-U.S. Border Area", Final Report, SCERP Project No.: AQ95-10, 1997.

5. Dworzanski, J.P.; Meuzelaar, H.L.C.; Maswadeh, W.; Nie, X.; Cole, P.A.; Arnold, N.S."Development of Field Portable Mass Spectrometric Techniques for Particulate OrganicMatter in PM-10," Proceedings of the 1993 International Symposium on Field ScreeningMethods for Hazardous Wastes and Toxic Chemicals, February 24-26, Las Vegas, NV,1993, 1, 517-541. Sponsored by the Southwest Center for Environmental Research andPolicy (SCERP).

6. Sheya, S.A.; Dworzanski, J.P.; Meuzelaar, H.L.C. “Characterization of Organic Constituentsin Tropospheric Aerosols by Novel Rapid GC/MS Techniques”, Air Waste and ManagementAssociation, 1999, in press.

7. McRae, G. J.; Goodin, W. R.; Seinfeld, J. H. "Development of a Second-GenerationMathematical Model for Urban Air Pollution - I. Model Formulation", AtmosphericEnvironment, Vol 16, No. 4, 1982, 679-696.

8. Tejeda, D.; Mejia, G.: "Impact of NAFTA on Mobile Source Emissions along theDallas-Saltillo Corridor", Proceedings of the International Symposium on EnvironmentalEngineering and Earth Sciences, October 26-30, Cholula, Puebla, Mexico, 1998, in press.

9. Kishan, S.; Meredith, R.; Weyn, C. "Development of Mobile Emissions Factor Model forCiudad Juarez, Chihuahua", Air Quality Planning Division (AQP), Texas Natural ResourceConservation Commission, Austin, Texas, August 30 1996.

10. Mendoza, A. "Aplicacion Preliminar del Modelo Fotoquomico de Calidad del Aire CIT a laZona de la Frontera Mexico-Estados Unidos", MS Thesis, ITESM, Monterrey, N. L., August1996.

11. Seinfeld, J. Atmospheric Chemistry and Physics of Air Pollution, John Wiley and Sons, NewYork, 1986.

12. Mukerjee, S.; Shadwick, D.S.; Dean, K.E.; Carmichael, L.Y. Assesing Transboundary Issuesin the Lower Rio Grande Valley, 92nd Annual A&WMA Meeting, St. Louis, June 1999.

13. Rogge, W.F.; Hildemann, L.M.; Mazurek, M.A.; Cass, G.R. and Simoneit, B.R.T. "Sourcesof Fine Organic Aerosol, 1-9" Environ. Sci. Technol., 1991, 24, 1112-1125 (part 1) thru1998, 32, 13-22 (part 9).

Table 1. Average Particle Size Distributions at the monitoring sites in the LRGV.

Hidalgo, Texas. March 11, 1998.Sampling period: 10:30 to 18:30 hrs.

dp(microns) N (dp)Vi (thousand of

microns3)Vi/Vt ΣVi/Vt

0,3-0,5 801026 26.84 0.0482 0.04820,5-1,0 124480 27.50 0.0494 0.15501,0-2,5 20935 58.75 0.1056 0.20332,5-5,0 7061 194.95 0.3504 0.5536

5,0-10,0 1124 248.35 0.4464 1.0000Total 954626 556.39 1.0000

Matamoros, Tamps. March 12, 1998. Sampling period: 0:00 to 4:30 hrs.

dp(microns) N (dp)Vi (thousand of

microns3)Vi/Vt ΣVi/Vt

0,3-0,5 889181 29.80 0.0374 0.03740,5-1,0 130958 28.93 0.0363 0.07381,0-2,5 28517 80.02 0.1005 0.17432,5-5,0 13737 379.31 0.4766 0.6509

5,0-10,0 1258 277.88 0.3491 1.0000Total 1063652 795.94 1.0000

Reynosa, Tamps. March 13, 1998.Sampling period: 0:00 to 2:00 hrs.

dp(microns) N (dp)Vi (thousand of

microns3)Vi/Vt ΣVi/Vt

0,3-0,5 1377389 46.16 0.0233 0.02330,5-1,0 469544 103.72 0.0523 0.07561,0-2,5 62086 174.22 0.0879 0.16352,5-5,0 36804 1016.22 0.5126 0.6761

5,0-10,0 2908 642.25 0.3239 1.0000Total 1948730 1982.57 1.0000

Reynosa, Tamps. March 13, 1998.Sampling period: 6:00 to 12:00 hrs.

dp(microns) N (dp)Vi (thousand of

microns3)Vi/Vt ΣVi/Vt

0,3-0,5 653962 21.91 0.3075 0.30750,5-1,0 47599 10.51 0.1475 0.45501,0-2,5 2420 6.79 0.0953 0.55032,5-5,0 520 14.35 0.2013 0.7516

5,0-10,0 80 17.70 0.2484 1.0000Total 704580 71.27 1.0000

Table 1. Continued...

Río Bravo, Tamps. March 13, 1998.Sampling period: 16:00 to 18:30 hrs.

dp(microns) N (dp)Vi (thousand of

microns3)Vi/Vt ΣVi/Vt

0,3-0,5 1538943 51.57 0.0722 0.07220,5-1,0 629158 138.98 0.1946 0.26681,0-2,5 37294 104.65 0.1465 0.41332,5-5,0 5736 158.38 0.2217 0.6350

5,0-10,0 1180 260.71 0.3650 1.0000Total 2212311 714.29 1.0000

Brownsville, Texas. March 14, 1998.Sampling period: 9:00 to 13:00 hrs.

dp(microns) N (dp)Vi (thousand of

microns3)Vi/Vt ΣVi/Vt

0,3-0,5 1026347 34.39 0.0175 0.01750,5-1,0 225479 49.81 0.0253 0.04281,0-2,5 81753 229.41 0.1165 0.15932,5-5,0 30711 847.97 0.4306 0.5899

5,0-10,0 3655 807.47 0.4101 1.0000Total 1367945 1969.05 1.0000

Hidalgo, Texas. March 14, 1998.Sampling period: 15:00 to 17:00 hrs.

dp(microns) N (dp)Vi (thousandof microns3)

Vi/Vt ΣVi/Vt

0,3-0,5 1236071 41.42 0.0229 0.02290,5-1,0 229274 50.65 0.0281 0.05101,0-2,5 72259 202.77 0.1123 0.16342,5-5,0 32180 888.55 0.4923 0.6557

5,0-10,0 2813 621.47 0.3443 1.0000Total 1572598 1804.87 1.0000

Table 2. Distribution of the sampling sites and the samples taken at each station.

Sample # MonitoringStation Start time End time GPS

1 Hidalgo, TX 10:15 am 12:15 pN: 26º 05’ 30.5’’W: 98º 16’ 17.0’’

2 same 12:21 pm 2:21 pm same3 same 2:52 pm 4:53 pm same4 same 5:30 pm 7:30 pm same

5Industrial ParkValle Hermoso

Tamps.11:45 am 1:45 pm

N: 25º 52’ 44.4’’W: 97º 33’ 25.5’’

6 same 1:50 pm 3:50 pm same

7 Reynosa, Tamps 11:00 am 2:00 pmN: 26º 03’ 44.8’’W: 98º 19’ 17.2’’

8 same 6:30 pm 8:30 pm same9 same 8:30 am 11:30 am same

10Rio Bravo,

Tamps.Downtown

4:25 pm 6:25 pmN: 29º 59’ 15.8’’W: 98º 05’ 46.5’’

11 Brownsville, TX 9:15 am 11:15 amN: 25º 53’ 55.5’’W: 97º 29’ 49.2’’

12 Hidalgo, TX 3:20 pm 5:20 pmN: 26º 05’ 45.8’’W: 98º 16’ 70.0’’

Table 3. Number of total and fecal coliforms as well as the presence of other microorganisms.

Sample# Monitoring Station

Total Coliforms(UFC/2.5 m3)

Fecal Coliforms(UFC/2.5 m3) Fungi

1 Hidalgo, TX 3 1 Aspergillus2 same 1 1 none3 same 3 2 Fusarium4 same 1 0 Alternaria sp.

5Industrial ParkValle Hermoso

Tamps.4 2

6 same 2 0 none7 Reynosa, Tamps 3 1 Penicillium8 same 2 1 Fusarium9 same 1 1 Alternaria

10Rio Bravo, Tamps.

Downtown3 1 Aspergillus

11 Brownsville, TX 2 1 Aspergillus12 Hidalgo, TX 1 0 none

Table 4. Factor Loadings after Varimax Rotation in Hidalgo, 1995 study.

Parameters Factor 1 Factor 2 Factor 3 Factor 4 Factor 5

4-hour daytime interval -0.376 -0.055 0.217 -0.320 0.216ambient temperature -0.099 -0.942 0.045 0.100 -0.136ambient pressure 0.209 0.205 -0.071 0.218 -0.121wind speed 0.371 0.030 -0.131 0.107 -0.710N/S wind vector -0.362 -0.177 -0.083 -0.066 0.050E/W wind vector -0.081 -0.193 -0.081 0.168 -0.910PM10 concentration -0.229 0.102 0.125 -0.949 0.087PM2.5 concentration -0.210 0.121 0.118 -0.955 0.061D PM (PM10 - PM2.5) -0.397 -0.130 0.178 -0.672 0.361åm/z 83-85 alkanes/alkenes (f) -0.500 -0.011 -0.017 -0.846 -0.009m/z 99 tributylphosphate (f) -0.196 -0.910 0.088 -0.039 0.084m/z 129 quinoline -0.165 0.073 -0.348 0.065 0.731m/z 149 DEP (f) 0.163 0.092 -0.945 0.092 0.121m/z 149 DBP (f) 0.027 -0.920 -0.071 0.161 -0.194m/z 149 DPP (f) -0.075 -0.964 0.124 0.108 0.007m/z 149 DOP (f) 0.013 -0.145 0.058 0.069 -0.087m/z 151 4-acetyl-2-methoxyphenol (f) -0.099 -0.930 0.124 0.088 0.009m/z 153 4-acetyl-2,5-dimethoxyphenol(f)

0.338 0.205 -0.542 0.332 -0.179

m/z 165 4-vinyl-2,6-dimethoxyphenol (f) 0.145 0.088 -0.922 0.188 -0.204åm/z 191 17a(H),21b(H)-hopanes (f) -0.917 -0.136 0.085 -0.243 -0.009m/z 191a hopane [C29H50] (f) -0.916 -0.037 0.073 -0.335 -0.005m/z 191b hopane [C30H52] (f) -0.892 -0.216 0.092 -0.159 -0.013åm/z 202 PNAHs [C16H10] -0.930 -0.053 0.150 -0.210 0.114m/z 202a fluoranthene -0.830 -0.018 0.046 -0.456 0.145m/z 202b pyrene -0.931 -0.076 0.218 -0.001 0.080m/z 219 retene (f) 0.131 0.060 -0.957 0.083 0.079m/z 239 methyl dehydroabietate (f) -0.092 0.139 0.112 -0.972 0.036m/z 306 tetraphenylene -0.128 0.136 0.088 -0.975 -0.012

(f) fragment ion, DEP = diethyl phthalate, DBP = dibutyl phthalate, DPP = dipentyl phthalate,DOP =dioetyl phthalate

Table 5. Measured and predicted NOx concentrations in the Highway from Brownsville toMcAllen / Hidalgo on March 14, 1998.

SITE HOUR Measured (PPB) Simulated (PPB)NO NO2 NOx NO NO2 NOx

1 12:28 7 13 20 12 74 862 12:45 15 22 37 2 16 183 13:05 69 40 109 2 14 164 13:21 93 27 120 2 4 65 13:40 115 50 165 1 5 66 13:56 43 35 78 1 10 117 14:18 12 21 33 3 23 268 14:26 27 28 55 3 23 26

� � �� � � �

� � � � � � �

�

� � � � � � � � � � � � � � �

� � � � � � � �

� � � � � � � � � � � � � � � � � � � � �

� � � � � � � � � � � �

� � � � �

� � � � �

� � � � � �

� � � � � � � � � �

� �

� � � �

� � � � �

� � � � � � �

� � � � � � �

� � � � � � � �

� � � � � �

� �

� � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � �

� � �

� � � � � � � � � �

� � � � � � � �

� � � � � � � � � � � � � � � � � � �

� � � � � � �

� � � � � � � �

� � � � � � � � � � �

� � � � � � � �

� � � � � � � �

� � � �

� � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � ! � � � � � � � � � � � � � " � � �� � � � � � � � � � � � � # � ! � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � # � � � � � � � � � � � � �

Figure 5b. Total ion profile and selected ion chromatograms illustrating the type of informationobtainable by direct thermal desorption (TD) GC/MS analysis of fine particulate mattertrapped on a quartz fiber filter. As discussed in more detail in the attached monographpreprint, the TD-GC/MS technique requires only microgram quantities (thus enabling time-resolved analysis at 1-2 hr intervals) rather that the milligram quantities needed by solventextraction (SX) GC/MS techniques (thus requiring 24 hr hi-vol sample collection). Asdeveloped in our laboratory, however, the TD-GC/MS technique produces very similarinformation on FPM adsorbates. Moreover, by going to pyrolysis ($500 C) rather thandesorption (250-350 C) temperatures information can also be obtained about high MWorganic FPM components (e.g. lignocellulosics, kerogens, synthetic polymers,microorganisms) and some thermolabile inorganic salts (nitrates, carbonates, sulfates).

Fa

cto

r Sco

res

/Sta

nd

ard

Dev

iatio

n

-1

0

1

2

3

-1

0

1

2

3

4

-1

0

1

2

3

4

SiteTime (hr)

12 24 36 48

-1

0

1

2

3

MIDNIGHT

NOONNOON

a) PC #4 "Urban Dust"

b) PC #1 "Automotive Emissions"

c) PC #3 "Biomass Burning"

d) PC #2 "Waste Burning"

Figure 11: Standarized principal component scores (varimax-rotated)corresponding to the loadings of the first four components inTable 2. Note characteristic morning and afternoon rush hourpeaks “in automotive emissions” component and major“urban dust” event in the late evening of the second day.

0

10

20

30

40

50

12/3/980:00

12/3/989:36

12/3/9819:12

13/3/984:48

13/3/9814:24

14/3/980:00

14/3/989:36

14/3/9819:12

Date, hour

SO

2 C

on

cen

trat

ion

, pp

bHIDALGO REYNOSARIO BRAVO BROWNSVILLEMATAMOROS

Figure 12. Predicted SO2 concentrations with the CIT photochemical model in the LRGV for March 12 to 14, 1998.

0

20

40

60

80

100

120

140

160

180

12/3/980:00

12/3/989:36

12/3/9819:12

13/3/984:48

13/3/9814:24

14/3/980:00

14/3/989:36

14/3/9819:12

Date, hour

O3 C

on

ce

ntr

ati

on

, p

pb

HIDALGO REYNOSARIO BRAVO BROWNSVILLEMATAMOROS

Figure 13. Predicted Ozone concentrations with the CIT photochemical model in the LRGV for March 12 to 14, 1998.

0

100

200

300

400

500

600

13/3/980:00

13/3/986:00

13/3/9812:00

13/3/9818:00

14/3/980:00

14/3/986:00

14/3/9812:00

14/3/9818:00

15/3/980:00

Date, hour

NO

x C

on

cen

trat

ion

, pp

bHidalgo

Measured, Hidalgo

Reynosa

Measured, Reynosa

Rio Bravo

Measured, Rio Bravo

Brownsville

Measured, Brownsville

Matamoros

Figure 14. Predicted NOx concentrations with the CIT photochemical model and actual measurements in the LRGVfor March 13 to 14, 1998.