WPCg 2 B?P Z CG Times 12ptphoenix#C\  P6QP#HP DeskJet 500HPDES500.PRSo\  PCxX,,0hXP2H< L/#|8"mo8;^=]8{{]8("SS]88]]ff.S8bulletE@6/\t M#=($*$W;E.E` ` ` 2+>g;;<f=";E@6/\Small Circle# =($E*$W;E.E"cover 2''@+#2p}wCЃ##o\  PCXP#level II0W=#2p}wCε# < #o\  PCXP#level IIIling, w0N=#:x2p}wCnX# < #o\  PCXP#2@]>??O@runin' @'#"s2p}wCV?##"s2p}wCV?#cover 1''!#2p}wC l## o\  PCXP#cover 3''!#2p}wC ##o\  PCXP#toc *O*#2p}wC l## o\  PCXP#2C -A!A"B#2Cozone *O*#2p}wC l## o\  PCXP#addressngs Summa!''3#2p}wCε##o\  PCXP#author subh"'4o#:b"zpCX# 5< #o\  PCXP#level Iddx#3d@q#2p}wC # ` #o\  PCXP#2I$/D%D&fG'FIlevel IVroup in $'z4V#:bLb:xЀC X# < # o\  PCXP#cover%Re-' 6 y 6dddy  6  6x y ^dddy #^2p}wC]#  n y  dddy  6  x y 6 dddy #^2p}wC]#author&' H yHdddy <  #:x2p}wC.}X# #:x2p}wC.}X#toc sub''@Y!'#"s2p}wCy?##"s2p}wCy?#2.M(J)J*K+Llevel 1('# Rm'#:x2p}wC.}X# ` #:x2p}wC.}X#title)=(O0'#2p}wC߼#- x  - #2p}wC߼#strategy*3|~@x#2p}wCa##o\  PCXP# x table txt+'z9|'# i2PkC &P## i2PkC &P#2W `MIPZQcdT"mo8;^tbttYtkYbttttb`(`lC2CC!CCCCCCCCCCd8YYYYYYzYzYzYzYC8C8C8C8ddddddddddYdddddodYYYYYYYdzYzYzYzYddddddddC8C8C8C8Ndz8z8z8z8z8ddddddCCCoNoNoNoNz8z8z8dddddddzYzYzYdz8dCoNz8dddddNF2[dCYddddd7>d<d<$8YYdCCddooCYȾd<d<+8oodCCddddCoȾ<чnn8!BBnnnyyP7c1RyyXyycnnnѐ~nyRzczXzcyhCBnndhcnnonvyXzXshn~XyBBnss~|y~~~~~~~~~~~~~~~~~~~XXXXXXXyyyyyyyyyyyyyyyyyyyyBBBBBBBBBBBBnnnnnnnsssssssssssszCCn"m^2CoddȧCCCdr2C28ddddddddddCCrrrdzNdzoȐC8CtdCdoYoYCdo8Co8odooYNCodddYO,OhC2CC!CCCCCCCCCCo8dddddȐYYYYYN8N8N8N8oddddooooddoddddzodddYYYYoYYYYddddddooN8N8N8N8do88888ooooddȐYYYoNoNoNoNCCCooooooȐdYYYo8oYoNCddodoNF2ldCddddddYJ 12p}wC Lq%V2*ܚV2PkCP'd8. d2p}wCNo] Ѷ2p}wCt,Y5(3\Y\  PCP7oC2o\  PCXP /xC8 Q:x2p}wCXW!C(ShAC\  PChP<5nC2%n*f9 xCXX7tC28t4  p(ACXyRzczXzcyhCBnndhcnnonvyXzXsh~XyBBnss~|y2g X K Ԋcover 1#2p}wCA#Ozone Transport Assessment Group'!cover 1#,8 Pk;,P# [  Xn L:L:cover 2#2p}wCA#Air Quality Analysis Workgroup%@+cover 2#,8 Pk;,P# [  X Qcover 1#2p}wCA#Draft Meeting Summary !cover 1#,8 Pk;,P# [  XJ Ocover 3#2p}wCA #Waterside Marriott Hotel SNorfolk, Virginia  X RoSeptember 25, 1996!cover 3#,8 Pk;,P# [ [0*0*0*  Y [#o\  PCWXP#n`(#(#ă  H= #V2PkCP#DRAFT Meeting SummarySeptember 25, 19963W of 14`J (#October 10, 1996ڑ X #d2p}wCA#`(#Air Quality Analysis Workgroup   yxdddy 4   J*|~strategy#2p}wCA#OTAG Air Quality Analysis Workgroupj*xstrategy#Xp Prk;WXP# x ۃ 44The Air Quality Analysis (AQA) Workgroup met on September 25, 1996, at the Waterside Marriott Hotel in Norfolk, Virginia. AQA Workgroup CoChair David Guinnup of the U.S. Environmental Protection Agency (EPA) convened the meeting and said its purpose was to develop consensus points about the workgroup's technical conclusions and recommendations and to identify those areas where significant differences of opinion exist within the workgroup. During the meeting, the following presentations were made:  YQ X` hp x (#%'0*,.8135@8:6(Rmlevel 1 ` # o\  PCXP# Because OTAG is more of a regional, and not an urbanoriented, ozone group, the different dimensions of the model's performance outside the urban areas need to be considered. One question is whether the modeling is sound enough to do what OTAG has to do. Dr.Husar said that this question is difficult to answer. If the issue is regional transport into the northeast region, and the model underpredicts in the northeast region and does not therefore detect accurately the scale of the problem, then it may be difficult to come up with reasonable scenarios. He said there is a need to establish the channels of communication on the efficacy of the model; this is a topic for further workgroup discussion. 8 This kind of analysis is necessary, but not sufficient, to gauge the model's effectiveness, and other kinds of data need to be considered. The question is how the model will eventually respond to all the control scenarios; as of now, only the basecase can be considered. In general, the features of the model seemed to represent adequately the air quality data; the model's relative lack of representation of specifics may not impair the future ability to use the model to get directional insights about what might happen in using one control strategy versus another. Potential turf issues arise when analyzing the model's results; a question is whether this workgroup should be involved in independent evaluation of model results, which are now becoming widely available. Additional diagnostic analysis is ongoing, and modelers will welcome further analysis. Directionality is one issue; another issue is whether the model performance will be good enough to identify specific sources to be controlled. There may be problems with vertical mixing; this issue is currently being considered in the southeast region. The input from thesesessions is being used to help solidify understanding and to interpret results moreappropriately.  P (#level 1#:x2p}wC QX#Report on Ozone and Tracer Data Analysis>(Rmlevel 1 ` # o\  PCXP# Eric Edgerton, of Environmental Science & Engineering, Inc., presented additional information on the ongoing study of precursor relationships at sites and the attempt to classify  X! ozone episodes based on the cobehavior of various pollutants, including NOy, CO, and SO2. NOy has been demonstrated to covary rather substantially with ozone concentrations. Mr. Edgerton discussed the analysis done for the Arendtsville, Pennsylvania, sitethe one  X% location within the Narstone Northeast network that had continuous NOy, CO, SO2 and ozone measurements during 1995. Most of the data was from the Narstone Northeast 1995 field season; a portion of the data was provided by EPA. This material, he advised, has not yet been reviewed by either EPA or the Narstone Northeast community. 8)/**ԌThe Arendtsville site is in southcentral Pennsylvania, approximately 120 kilometers west of Philadelphia and 120 kilometers northwest of Baltimore. The observation site is on geographically high ground relative to the area and thus receives good fetch"an important component to making sound measurements in the vertical dimension of the atmosphere of the  X` tracers of interest, such as nitric acid and SO2, which deposit very quickly. At this site, a fairly strong relationship had been observed between ozone and NOy during the midafternoon hours. The initial data was a time series of 90 dailymaximum, 8hour rolling averages of ozone concentration. Tracer information was sought for those days that recorded an average of 80 ppb of ozone. Significant differences were seen in sourcesignature ratios  X for CO and NOx, as well as for SO2 and NOx. Automobiles have a ratio of CO to NOx at their tailpipe of 8 to 1; for point sources, the ratio of CO to NOx is quite lower. Using these ratios, Mr. Edgerton noted, afforded some degree of diagnostic ability to differentiate between sources. These ratios are affected by the atmospheric lifetimes of their components; for example, CO has an atmospheric lifetime of many days and therefore does not disappear from  Xh the atmosphere as rapidly as NOx and SO2, which have comparatively much shorter lifetimes. The classification scheme defined o44An urban source with a high COtoNOy ratio in the air mass.(#4  X o44A major point source with a high SO2ĩtoNOy ratio.(#4 o44A mixed category of sources with both. (#4 This information was then applied to the database collected, and each day that had high ozone concentrations at the Arendtsville site was classified. To definitize the results, background CO in the atmosphere was accounted for by subtracting it from the ratios. They generated a time series of many different types of days21 in all. Mr. Edgerton discussed in detail a 3day episode, July 1214, 1995, which encompassed the highest ozone concentration noted at the site during the 1995 OTAG episode. During this entire period,  X high CO levels were noted, with varying levels of SO2. July 12 was classified as an urbaninfluenced day, as most of the NOy appeared to be associated with a carbon monoxide  X# signal. During the next 2 days, the amounts of SO2 increased, albeit erratically, and both days were therefore classified as mixed, with contributions from both source categories. In general, of the 21 days observed, 11 were classified as mixed, with contributions from  X`' both elevated CO and SO2 concentrations; 8 were classified as urban, with contributions  Xx( primarily from elevated CO, with smaller amounts of SO2; and 2 appeared to have most of  X) the NOy associated with NO2 sources, with little elevated CO concentrations. These results to)/** some extent confirm what was coming out of the model. Not only was there a good deal of quantitative agreement between the model and the measurement, Mr. Edgerton said, but there also appeared to be some qualitative agreement as well because the model does suggest urbaninfluence. Mr. Edgerton summarized the findings as follows: o44Similar relationships were noted between ozone and NOy, as noted elsewhere. (#4 Ho44From the tracer data, similar mixes of sources were noted at Arendtsville as at the three SOS sites operated in the southeast region, with a breakdown in apparent contribution among urban, point, and mixed sources.H(#4 o44The mixedsources category was slightly larger numerically than the urbansources category, with the pointsources category ranking well behind.(#4 o44Work is ongoing to expand this analysis to include meteorological factors, other years and other sites (to the extent that data is available), and expanded chemistry (such as NOz) to afford a more definitive example of the chemistry occurring atsites. (#4 Ozone has a definitive signature and seems to be very plumy in its behavior. Downward  X mixing of elevated SO2 occurs frequently and, if that had happened during the maximum ozone concentration incidence, that circumstance would have been reflected by the 8hour average corresponding to the maximum daily rolling average that was used to determine the contribution of sources. The return of nitric acid at the surface also occurs often. A written summary of this analysis will be posted on the Web site. Mr. Guinnup mentioned that 20 copies of the draft summary of technical reports are available for participants to look at, including a list of all the technical reports and a summary table. The list was not inclusive or final, he advised. He also mentioned that a literature review of some airquality analysis products is included on the list, and he asked participants to add to that list as they see fit.  PH& (#level 1#:x2p}wC QX#Report From the Air Trajectory Analysis GroupW(Rmlevel 1 H&` # o\  PCXP# Rich Poirot of the Vermont Department of Environmental Conservation summarized the status of this longterm project. Using data from measurements taken four times a day for seven summers, they have currently run back trajectories for 23 monitoring sites. They have)/** recorded results from six sites at higher elevations in the Appalachian Mountains. Also, they have added six southern sites and 12 sites along the eastern seaboard and upper midwest regions. Their basic analytical technique is to run a large number of back trajectories from a site, keeping track of their spatial and temporal characteristics on a grid of 1,440 80by80 kilometer squares, which has the effect of truncating the average trajectory length after distances reached approximately 72 hours out from the source location. Ozone metrics used for the monitoring sites have changed during the analysis, Mr. Poirot continued. At the higher elevation sites, which had little diurnal variation, they used raw ozone data and did not worry about when the measurements were made. At the lower elevation sites, however, which exhibited much more pronounced diurnal variation, they calculated the diurnal average at each hour of the day, to make the data comparable at different times during the day; they then expressed the ozone metric as the deviation from thediurnal mean. In combining data from different sites, they aimed to standardize the data to make it as comparable as possible, and then used geometric means and standard deviations to express the ozone data in terms to answer the question about how many geometric standard deviations above or below the mean is the concentration. In other words, he said, they aimed to determine whether the ozone concentrations were high or low, and how high or low theywere. The analysis revealed a lot of spatial coverage over much of the domain. Their basic technique had been to evaluate trajectories to arrive at a receptor. The approach had been to sort the trajectories on the basis of the ozone concentrations at the receptor, to consider each grid square when sorting the trajectories, and thereby to arrive at the receptor according to observations of whether or not the trajectories had passed through the grid squares. Most previous results had been based on this locationbased sorting. One advantage to this approach was that it used all available trajectories and therefore accommodated a fairly robust data set and generated results in familiar units. A disadvantage of the locationbased approach is its weakness at characterizing episodes and addressing nearreceptor influences. These results, Mr. Poirot emphasized, were taken from a concentrationbased sorting approach, which categorized the metrics at the receptor into bins and examinedbecause of levels of ozone noted at the receptorwhere that air had previously resided. This approach can be applied to the entire distribution of ozone data, including the episode periods, and it also allowed for a convenient spatial comparison among different ozone subsets. When applying this technique, however, the measure obtained was in units of time and not in ppbs of ozone. Another disadvantage of this technique, as originally applied, was that it tended to overstate probabilities near or at the receptor. To avoid this, the technique has been) /** geometrically adjusted by first calculating an everyday probability field for air parcel origin for each receptor location and then investigating where the air most likely would have been if the ozone levels were high. `Mr. Poirot then presented some of the results for locations associated with belowaverage concentrations: a group of midwest sites, a group of northeast sites, a group of southeast sites, and a group of all the sites aggregated. Some of these observations included thefollowing:` o44In most cases, there did not seem to be extreme differences between a very high and moderately high percentile. (#4 o44There may have been a smaller area associated with higher concentrations ofozone.(#4 o44At southern sites, at any percentile level, a substantial increase in residence time probability occurred around the monitoring site during high subset days, suggesting that stagnation around the monitoring site is a more significant factor in the southeast region.(#4 The extent of seasonal influence on the differences in the patterns that had been observed was also examined. Using data from the work conducted by S.T. Rao of the New York Department of Environmental Conservationwhich statistically disaggregated ozone data into synopticscale, longterm trend, and seasonal componentsthey took the shortterm component and conducted a similar analysis, removing the seasonality and longterm trend components from the raw ozone data. The similarities between Dr. Rao's data and their raw ozone data were striking, Mr. Poirot said, and provided indications that seasonal deviation was not the cause of the differences in the patterns. Mr. Poirot suggested the following conclusions from this work: o44OTAG is geographically a wellformulated region for examining ozone transport. Virtually all the clean air originates from outside the region, and most high concentrations originate internally. (#4 o44Very few locations within the OTAG domain can claim they are never upwind of some other site when ozone concentrations are above average; the only exceptions may be some easternmost peninsular regions. (#4 ) /**Ԍo44With a few exceptions, no substantial differences occur in site patterns between areas upwind from moderately high ozone concentrations and areas upwind from extremely high ozone concentrations. Exceptions include the Shenandoah and Great Smoky Mountains, where site patterns for extremeepisodes are very different from those of moderately high days. Higher elevation sites are very heavily influenced by transport; no matter where these sites are located, higher elevation sites tend to have longer transport distances when using this method. High elevation sites also have substantially lower maximum concentrations than adjacent, lower elevation sites. (#4 o44With a few exceptions, the relative influence and distance of transport tends to increase and be magnified as one moves from southwest to northeast in the OTAG domain. With southern locations, areas of high residence time probability tend to congregate in relatively small areas near the monitoring site; with northeastern locations, fairly high incremental probabilities often occur at fairly long distances from the monitoring site. (#4 o44Each individual source region has areas unique to that site. Alternatively, some areas are commonly upwind from many monitoring sitesareas that tend to be centrally located in a region and where there are high and persistent stationarysource NOx emissions present. It is still difficult to characterize the relative influence of a central location when compared to the relative influence of having emissions that contribute.(#4  o44When using multiple sites, it is important to investigate the distribution of data around those sites so that one extreme Zscore does not disproportionately affect the analysis.(#4  P (#level 1#:x2p}wC QX#Report on Additional Trajectory Analysis Work~v(Rmlevel 1 ` # o\  PCXP# Dr. Rao defined the object of this work as using the patternrecognition capabilities of the clusteranalysis methodcoupled with trajectory modelsto relate synoptic meteorology to pollution climatology at a given site. The clustering methodology, he said, is unique, straightforward, and able to analyze vast amounts of data. In their analysis using this methodology, they looked at clustersegregated ozone data, as well as other pollutants such as zinc, arsenic, and selenium; analyzed that data statistically; and extracted the analysis results pertaining to synoptic meteorology and pollution climatology. The study included calculations of 3day backward trajectories. In this case, Dr. Rao noted, they used the Whiteface Mountain location. A total of 612 trajectories were analyzed over) /** the 5year study period. Dr. Rao advised those who are curious about clustering methodology  X to refer to an article in a 1992 edition of Atmospheric Environment. Clustering methodology, he said o44Selects and seeds trajectories.(#4 o44,` ` 44Computes the real trajectories.(#` """""""" o44Determines how close the two are to one another.(#4 """""""" o44Recalculates the average trajectory depending upon the distance between the real and the seed trajectory.(#4 o44Recalculates the cluster and the deviation, and reduces the clusters to see if the deviation has changed.(#4 The example Dr. Rao presented included 30 clusters of 612 trajectories. Analyzing the trajectories independent of the concentrations determined how many significant statistically different clusters the 612 trajectories could fit into. By using the methodology, they determined that only eight clusters were needed to meaningfully classify trajectories. In analyzing a cluster, they constructed a composite surfacepressure map for all the trajectories to identify a particular synoptic pattern prevailing for that particular cluster. They then used a statistical test to determine whether the clustering was robust and whether significant statistically different clusters had been generated by the clustering methodology. Once this clustering had been performed, the concentrations in each cluster were analyzed and the levels characterized. They then characterized the regions where the 3day back trajectories ended up and determined whether the mean concentration from a particular cluster for a particular species is statistically significantly different from the mean from a different cluster, to clearly distinguish the clusters and the 3day end points of the trajectories. For ozone concentrations, the following was revealed: o44A southeasttosouthoriented cluster (southeast to south being the spatial extent of the 3day trajectory) produced the highest amount of ozone at Whiteface Mountain.(#4 o44A westtosouthwestoriented cluster produced a medium level of ozone.(#4 """""""" o44A northtonortheastoriented cluster produced the lowest level of ozone. (#4) /**ԌEach cluster orientation was attributable to specific synoptic forces, Dr. Rao said. For trace species, only one cluster each was statistically significant for sulfur, zinc, and selenium. In conclusion, Dr. Rao said, the discrimination afforded by the clustering methodology coupled with trajectory analysis can be very useful in understanding the relationship between synoptic patterns and pollutantconcentration levels measured at a site. Specifically o44A number of synoptic patterns were identified, each of which was characterized by pollutant transport from a different source region.(#4 o44The results suggested that, for ozone, southtosouthwesterly transport led to significantly higher ozone concentration measurements at Whiteface Mountain.(#4 o44For trace species, northwesterly flows brought high concentrations of arsenic to the site, and southwesterly flows brought concentrations of zinc, selenium, manganese, and sulfur to the site.(#4 This methodology will continue to be explored, Dr. Rao said, to better understand the relationship between synoptic and associated flow patterns that lead to the buildup of concentrations of particular types of species elements. This kind of analysis is being extended to lower elevation sites as well. It is important to distinguish between ozone and the inert metals species; particle species are not changing. The model, although able to look at some threedimensional characteristics, may be relatively insensitive to changes that might affect the production or loss rates of species in vertical transport.  PX (#level 1#:x2p}wC QX#Discussion on Differences Between the Two Approaches:(Rmlevel 1 X` # o\  PCXP# The two approaches may not be compatible. Dr. Rao's approach suggested a strong tendency to look in the direction of the northeast corridor, as opposed to what was suggested by Rich Poirot's approach. The differences between the two approaches may just be a matter of degree; for almost all the northeast sites considered, a definite east coast corridor influence has been noted. A possible difference between these techniques is that Mr. Poirot's approach would not account for depletion along the trajectory as the material is passing from origin point to Whiteface Mountain, whereas Dr. Rao's approach allows for less time for any lateral diffusion and possible removal processes to occur. Mr. Poirot responded that there is no chemistry in either approach; the difference may be, he said, that his approach truncated the  X& trajectories at an average of 3 days, whereas Dr. Rao's approach truncated every trajectory at 3 days. ( /**ԌDr. Rao's approach grouped trajectories first and then looked at means, and it analyzed the average concentration for a given trajectory. Mr. Poirot's different sorting approach afforded more of an indication of the frequency of occurrence of a given trajectory. The sorting philosophies differed: Dr. Rao's approach sorted on the basis of trajectory; Mr. Poirot's, on the basis of concentrations and locations. A frequency component is missed by the clustering inherent in Dr. Rao's approach because any individual cluster can occur more than once. Dr. Rao's approach has some demonstrated strengths: it corresponds to meteorology, and it can discriminate on the basis of not only ozone, but also the trace elements. With these traceelements, however, there is some doubt over how this approach handles the problem of belowdetectable limits for these species. Dr. Rao admitted that the detectionlimit problem issignificant; one of his students is evaluating this topic. In response, they used extrapolationmethods.  PP (#level 1#:x2p}wC QX#Review of Work Products(Rmlevel 1 P` # o\  PCXP# Mr. Guinnup turned the meeting to a review of the draft of the summaries of individual work products. The draft did not represent a complete summary, he advised, and the individual drafts presented had been developed independently of the individual researchers and the workgroup. The list itself had not been placed on the Web site at the time of the meeting. He said the first two pages represented an attempt at listing and categorizing the individual papers, and he asked the meeting participants to review the list for inclusiveness. The next two pages took that information and placed it in tables to establish the written reports' availability and completeness, and Mr. Guinnup asked the participants to review that list for accuracy. Some of the reports and work products, he said, were clearly still in progress, including the observationbased modeling efforts. Some of these works in progress, Mr.Guinnup advised, will have an indeterminate impact on the workgroup's interpretations and consensus building. The handout also included summaries of some of the individual work efforts; these distillations of the salient features of these projects were done without the input of the study's individual researcher. Mr. Guinnup asked the participants to make sure these summaries appropriately captured the essence of the studies. At some point, workgroup members will need to summarize what they can and cannot do with their airquality analyses, he said, and appropriately characterize their findings to the Policy Group. &/**Ԍ P (#level 1#:x2p}wC QX#Model Performance and Transport(Rmlevel 1 ` # o\  PCXP# Mr. Guinnup asked the group to focus on developing consensus points on the topics of model performance and transport. The policyrelevant question regarding model performance is whether the models are performing well enough. He suggested breaking the workgroup down into four workshops, to address the following questions: o44Do the individual analyses support the concept of ozone transport, and if so, canitbe quantified? Is ozone transport a problem? Is ozone, or its precursors, actually transported? The answer to these questions will have to address the matter of scale.(#4 o44What aspect of ozone transport is most relevant to OTAG?(#4 o44How well does the model compare to the airquality picture over the entire OTAG region? Prior modeling performance guidelines have been based on urbanarea modeling, not regional modeling; alternatively, the workgroup needs to get the results from prior analyses of the performance of regionalscale models. (#4 o44Do any underlying biases alter the interpretation of modeling results? Some of the work presented during this meeting suggest the presence of these biases.(#4 A participant stated that the key issues may be whether the modeling episodes are representative. Mr. Guinnup agreed that the accuracy of the modeling episodes is an important issue, and one that has been answered"for example, by some of Dr. Husar's analyses, which have studied patterns of episodes against overall patterns. In response, a participant stressed the need to include in the list a report by modelers specifically addressing this issue. Because of time constraints, Mr. Guinnup asked the participants to reconvene after the subgroup's afternoon session.