This summary was prepared as an informational document for use by the OTAG Air Quality Analysis Workgroup. Two complementary assessments were performed and are summarized here. (1) Transport and mixing phenomena related to ozone exceedances in the northeast U.S. were examined using data from 1995 NARSTO-Northeast measurements. (2) Results of the OTAG modeling for the northeast U.S. were evaluated using available air quality data for 1988 and 1991 and using air quality and meteorological data from NARSTO-Northeast for 1995. Stand-alone reports were prepared for each of these assessments.

This summary is a working draft document and has not been peer reviewed or reviewed by the NARSTO-Northeast participants. All conclusions are those of the authors and not of NARSTO-Northeast or the sponsors.

The authors appreciate the cooperation and input of the NARSTO-Northeast participants. This summary and the underlying reports draw heavily on the NARSTO-Northeast database and on information presented at the First NARSTO-Northeast Data Analysis Symposium and Workshop at Norfolk VA in December 1996. We are especially grateful for data, analyses, and technical input from Bob Gaza and Doc Taylor of NYDEC, Bill Ryan of the University of Maryland, Jian Zhang of the State University of New York at Albany, and Paul Roberts, Jerry Anderson, and Tami Haste of Sonoma Technology, Inc. (STI).

This effort was initiated and coordinated by Jeffrey West of General Public Utilities (GPU) and was made possible by funding from GPU, New England Electric System (NEES), and the Electric Power Research Institute (EPRI).

The authors thank the U.S. EPA and the OTAG modeling centers for providing the UAM-V model results and routine observational data, and thank the NARSTO-Northeast sponsors for making the special study data available.

The OTAG model evaluation report is partially based on a separate report titled ìAssessment of UAM-V Model Performance in the Northeastî (Londergan and Moore, 1997) prepared by Earth Tech for a consortium of Northeast utilities.

The authors appreciate the hard work and long hours of Carolee DeWitt, Sandy Barger, and Martina Shultz of the STI publications staff who made it possible to complete these reports on a very short time schedule.

In the summer of 1995, NARSTO-Northeast measurements were performed as part of the North American Research Strategy for Tropospheric Ozone (NARSTO) to enhance the understanding of factors governing ozone formation and transport in the Northeast United States. NARSTO-Northeast provided an extensive database on the three-dimensional distribution and transport of ozone and precursors in the northeast US. This database has made possible (1) an assessment of the transport and mixing phenomena affecting ozone episodes in the Northeast Ozone Transport Region (OTR) and (2) a detailed evaluation of modeling performed in mid-1996 by the Ozone Transport Assessment Group (OTAG) for subregions of the OTR.

Analyses of NARSTO-Northeast data showed that transport of ozone and precursors is affected by flow regimes that differ with altitude. Near-surface flows typically were from the south to southwest, flows from 200-800 m msl were from the southwest to west, and flows from 800-2000 m msl were from the southwest to northwest. The aloft flows resulted in transport of ozone and precursors over distances of 300-800 km (200-500 miles) in 24 hours, often with the greatest velocities at night. The near-surface speeds were lighter than those aloft and near calm during late night hours, resulting in accumulation of emissions in the night and morning hours and subsequent transport distances of half or less of those aloft.

During the episodes studied, 30-50 ppb of excess ozone that was transported aloft overnight, in combination with an assumed non-anthropogenic background of about 40 ppb, resulted in measured regional background concentrations of 70-90 ppb or more over much of the OTR. If such conditions typically occur more than two days/year, this transport contribution might be sufficient to cause exceedances of the proposed new eight-hour 80 ppb federal ozone standard in some regions.

The above background amount is about half of the 150-180 ppb peak one-hour concentrations seen downwind of urban areas. If the chemistry were linear, simple superposition would suggest that clean air and transported excess ozone each might account for about 25 percent of the one-hour peak ozone downwind of urban areas, with the other 50 percent coming from same-day urban emissions. Simple box-model calculations, however, demonstrate that the chemistry is not linear and suggest that reducing the excess ozone aloft may not reduce the peak surface concentrations downwind of urban areas in proportion. Emissions from urban areas were seen to cause exceedances of the current federal standard 50-250 km downwind and could have done so even without a contribution from excess ozone aloft. In particular, ozone and precursors transported offshore near the surface from Massachusetts resulted in ozone concentrations of over 150 ppb at shoreline sites in Maine.

For the OTAG modeling of the July 14-15, 1995 ozone episode in the OTR, the speeds, directions, and transport distances of the simulated regional flows above 500 m agl were qualitatively similar to the measured flows, however details of mesoscale flows such as the night jet were not resolved. The near-surface flows were less accurately resolved, resulting in errors in the location of impact of urban emissions, especially near the coast. These results are not unexpected due to the coarse resolution of the OTAG models. The OTAG modeling was intended to capture the larger regional-scale features of ozone episodes and to characterize the relative contribution of regional versus local emission sources. It was not designed to simulate accurately the detailed concentration distributions near individual urban areas.

For the OTAG modeling for 1988, 1991, and 1995, the magnitude of the highest one-hour simulated concentration in a subregion of the Northeast was often comparable to the highest observed at any station in the subregion. However, the exact time and location of the simulated one-hour daily maximum concentrations were rarely accurate. This mismatch in space and time is partially due to inaccuracies in the simulated wind fields and is typical of air quality model simulations.

For most days, the comparisons of observed and simulated ozone concentrations in the Northeast Corridor showed little prediction bias for peak and mean concentrations in ìurban plumes.î The model error was comparable to typical urban-scale modeling results for many subregions and days. However, there were some days for which the modelís bias and error did not meet EPAís urban-scale photochemical modeling acceptance criteria, suggesting caution about the reliability of the modeling effort.

For each episode evaluated, the mean bias (throughout the model domain) in ozone showed a systematic trend with time during the episode (known as bias creep). The model consistently underestimated the observed concentrations for the first few days of the episodes and overestimated the observed concentrations for the last few days of the episodes. This systematic trend may indicate a problem with the model formulation or model inputs that may influence the accuracy of its simulated response to emissions controls.

Boundary layer ozone flowing into the urban corridor was consistently underestimated in the simulations. The daily maximum concentrations in the upwind areas typically were underestimated by 10-20 ppb. The model commonly underestimated the morning ozone aloft and, therefore, the carryover of ozone from the previous day. Comparisons of simulated and observed vertically integrated ozone concentrations show the model simulations typically underestimated ozone aloft by 10 and 5 ppb in the morning and afternoon, respectively. Therefore, the simulated and observed afternoon vertical profiles often seemed to show good agreement and little bias.

Comparisons of simulated and observed vertically integrated NO_y,, RHC, and CO concentrations show the simulations mostly overestimated NO_y concentrations and mostly underestimated RHC and CO concentrations. Based on all of the aircraft samples collected during the July 1995 episode, the model estimated 27 ppbC of RHC aloft when 90 ppbC was observed; however, the number of data points for this comparison was small.

Section Page

PREFACE AND ACKNOWLEDGMENTS ii

ABSTRACT iii

1. INTRODUCTION S-1TRANSPORT AND MIXING PHENOMENA THAT AFFECT OZONECONCENTRATIONS IN THE OTR S-3EXTENT THAT OTAG MODEL RESULTS REFLECT IMPORTANT TRANSPORTPHENOMENA S-7DIFFERENCES BETWEEN MEASURED AND MODELED OZONECONCENTRATIONS S-105. REFERENCES S-13

1. INTRODUCTION

The Ozone Transport Assessment Group (OTAG) was initiated by the Environmental Council of the States (ECOS) in 1995 to develop a consensus on how to deal with the transport of ozone and precursors in the eastern United States. Among the issues confronting OTAG is the scope of regional transport of ozone and its precursors: how to assess the influence of transport on ozone concentrations and to design emissions management strategies to account for it.

OTAG performed ozone modeling for the eastern United States to characterize regional transport and assess the effects of emission reductions (OTAG, 1996a). The modeling was performed for a large region extending from Florida to northern Michigan and from Texas to Maine. The scientific credibility of the OTAG modeling is derived from the formulation of the model, the quality of the input data, and the performance of the model in these specific applications. To evaluate model performance, OTAG provided a comparison of measurements and model simulations for the entire domain and for four large portions of the domain, including the Northeast (OTAG, 1996b). OTAG did not evaluate the model for subregions within the Northeast.

During the summer of 1995, a public-private consortium sponsored expanded measurements as part of the North American Research Strategy for Tropospheric Ozone (NARSTO) to enhance the understanding of factors governing ozone formation and transport in the Northeast United States. Dubbed "NARSTO-Northeast," this study provided an extensive database on the three-dimensional distribution and transport of ozone and precursors. Extensive surface and upper-air air quality and meteorological data were obtained for several ozone episodes in 1995 and 1996, and limited data were obtained in 1994. The 1995 observational campaigns are described by Roberts et al. (1995), and the database and data access procedures are described by Korc et al. (1996).

The NARSTO-Northeast database has made possible (1) an assessment of the transport and mixing phenomena influencing ozone episodes in the Northeast Ozone Transport Region (OTR) and (2) a comparison of the OTAG model simulations with measurements for subregions of the Northeast portion of the modeling domain. Complementary reports addressing these two topics have been prepared by Blumenthal et al. (1997) and Lurmann et al. (1997), and their results are summarized here. These assessments were intended to address the following objectives.

1. Identify important transport and mixing phenomena that influence ozone concentrations in the OTR.

1. Determine the extent to which OTAG model results reflect these phenomena.

1. Estimate the directional differences between measured and simulated ozone concentrations that might occur due to differences between measured transport phenomena and the processes accounted for in the simulations.

A series of questions was posed for each objective. The first objective was addressed by Blumenthal et al. (1997) and the third by Lurmann et al. (1997). Time was not available to perform specific analyses to address the second objective, but the questions for the second objective were partially addressed through examination of the results reported in the above two reports.

The following sections summarize the answers to the questions posed for each objective.

1. TRANSPORT AND MIXING PHENOMENA THAT INFLUENCE OZONE CONCENTRATIONS IN THE OTR

This section summarizes important transport and mixing phenomena identified during NARSTO-Northeast that were related to exceedances of the federal ozone standard during the 1995 measurement period. The results described are based on the ozone episodes of July 13-15 and July 31-August 1, 1995.

• What are the major transport regimes during ozone episodes?

Transport of ozone and precursors is influenced by flow regimes that differ with altitude. These regimes can be divided into three general categories plus a subset of the lowest-altitude regime.

1) Boundary Layer Synoptic Transport roughly 800-2000 m msl

- Transports ozone and precursors across the Appalachians - from western OTR and Midwest

- Flows are from west to northwest during most episodes

- Results in transport distance of 300-800 km (200-500 miles) in 24 hours

- Does not have strong diurnal component

2) Channeled Flows Below the Ridge Heights roughly 200-800 m msl

- Includes nighttime low-level jets (occurred on 6 of 9 1995 regional O₃ episodes)

- Includes "lee trough" flow caused by the interaction of the synoptic flow and the Appalachians

- Often transports substances from southwest along urban corridor, but other directions occur, including flow from the west through gaps in the Appalachians

- Results in transport distances of 200-400 km (125-250 miles) overnight or similar to synoptic flow over 24 hours

3) Near Surface Flows roughly 0-200 m msl

- Light winds in night and morning allow accumulation

- Fresh emissions and urban plumes move downwind and react during daytime

- O₃ aloft and aged precursors are entrained as the mixing layer deepens

- Transport is typically to the north through east along the urban corridor for 50-250 km (30-150 miles) by evening. Transport distances over 24 hours are typically less than half of those for the above flow regimes.

3a) Offshore Flows (subset of 3 above) roughly 0-200 m msl

- Light winds in night and morning allow accumulation onshore

- Accumulated urban emissions can transport offshore and react during daytime

- Transport is typically to the northeast through east: Boston to NH, ME; Philadelphia and NJ to Long Island; Baltimore/Washington across bay to DE

- Transport can be 200 km (125 miles) or more during daytime and evening

- Layer stays stable and thin over water, with minimal mixing and no added emissions

- Can cause high concentrations at shoreline; dilutes with mixing as transported inland

• How do transport speeds and directions differ between the surface and aloft?

During the episode periods studied, the near-surface flows typically were from the south to southwest, the channeled flows were from the southwest to west, and the synoptic flows were from the southwest to northwest. The channeled flow directions varied during the nights, with more southerly flow earlier in the night and more westerly flows later at night. The surface speeds were lighter than those aloft, resulting in transport distances of half or less than those aloft. Typically, the surface speeds were lightest at night, while the channeled flow speeds were highest at night.

During the episode nights studied, the first two flow regimes above became decoupled from the surface flow, and the channeled-flow regime accelerated. During the day, winds throughout the boundary layer become coupled, and wind speeds and directions can become similar through all three regimes. During the day, frictional effects can slow the winds aloft below the nighttime speeds. On most episode nights studied, a jet occurred in a layer just above the surface, where the speeds were higher than those above or below. On some nights, the winds aloft did not form a jet (i.e., they were not faster than the winds both above and below). They were, however, faster than the winds below. This is a normal occurrence on most summer nights. When the frictional effects are decoupled, the winds aloft will be faster than the surface winds. When the aloft winds are decoupled from the surface, ozone and precursors aloft can be transported without deposition to or input from the surface layer.

The wind speeds aloft also varied from south to north, with lighter winds and shorter transport distances in the southern portion of the OTR. Average boundary layer synoptic wind speeds during episodes are 5-10 m/s in the New York City area and about 4 m/s in the Washington, DC area.

• How far and at what concentrations can ozone or precursors be transported aloft between episode days?

Trajectory analyses and aloft ozone measurements show ozone can be transported 300-800 km (200-500 miles) in 24 hours across the Appalachians at altitudes of 800-2000 m msl and 200-400 km (125-250 miles) overnight along the corridor or through gaps in the mountains below the ridge heights at altitudes of 200-800 m msl. Over 24 hours, the transport distances for the 200-800 m msl channeled flows are similar to those for the synoptic flow above, but more of the transport occurs at night in the channeled flow regime.

The aloft transport can result in the widespread occurrence of overnight ozone carryover aloft at average boundary layer concentrations of 70-90 ppb and with layers of over 100 ppb. The resulting background ozone is a mixture of ozone transported from different locations at different altitudes. When mixed to the surface, this transport can lead to surface ozone concentrations of 70-90 ppb. This surface background contribution is roughly 30-50 ppb above clean air, assuming a clean-air value of about 40 ppb. This background carryover ozone can be caused by emissions or ambient ozone coming from 200-800 km (125-500 miles) away since the prior day.

From early morning ozone, NO_y, and VOC measurements aloft, long-range transport seems to contribute mainly ozone and aged precursors, but fresh primary NO_x emissions also can be transported overnight in these flows.

• How far are precursors that are emitted during the nighttime and morning transported in the near-surface layer during the day; at night? How much ozone do they add to the regional background

Same-day plumes (or pulses) of ozone downwind of urban areas are embedded in a regional background. The peak urban plume concentrations were up to 80-100 ppb higher than the nearby regional upwind or crosswind concentrations. Offshore transport of near-surface emissions was seen to cause downwind offshore and shoreline concentrations of over 150 ppb, which were 80-100 ppb higher than the concentrations higher aloft and at downwind onshore areas not impacted by the transported plume. This increased near-surface ozone was confined near the surface and did not mix with the lower concentrations aloft until it was advected onshore.

Daytime transport distances (0500-1900 EST) of these near-surface emissions were in the range of 50-250 km (30-150 miles). During the night, surface winds were generally light and transport distances were on the order of 25-75 km.

Under these conditions, even with clean air as a background, urban plumes resulting from same-day emissions could cause ozone concentrations to exceed the one-hour federal standards 50-250 km (30-150 miles) downwind.

• How do mixing heights vary between episode and non-episode days? What is the effect of the difference, if any?

Mixing heights increased more slowly on widespread episode days. The lower morning mixing heights can result in accumulation of higher precursor concentrations in the surface layer in the morning. They will also result in higher precursor concentrations in layers transported offshore in the morning.

• What is the relative importance of long-range ozone and precursor transport for downwind ozone concentrations? How do ozone layers aloft affect surface concentrations?

Ozone aloft can influence surface concentrations through addition of aloft ozone and precursors to the surface concentrations as the mixing layer deepens and entrains them.

During episode conditions in the OTR, ozone transported aloft combined with the clean air background of 40 ppb can provide regional background concentrations of 70-90 ppb or more. This amount by itself may be adequate to cause exceedances of the proposed new federal standard; however it is only about half of the peak concentrations seen downwind of urban areas. This would suggest that if the chemistry were linear, clean air and transported ozone each might account for about 25 percent of the peak ozone downwind of urban areas, with the rest coming from same-day urban emissions.

To account for the chemistry of the interaction between ozone transported at various levels aloft and fresh surface emissions, simple two-layer box-model simulations were performed for conditions similar to August 1, 1995 and with many simplifying assumptions. Assumed urban emissions were varied to assess sensitivity to emissions, and two different chemical mechanisms were used. These simulations indicated that ozone carryover aloft might increase the maximum surface concentrations downwind of major urban areas by less than one third to over 100 percent of the aloft excess over 40 ppb (clean air) for the range of emissions used. Most of the simulations were in the bottom half of this range. On August 1, 1995, ozone carryover aloft was roughly 70-80 ppb. Given this range for the initial concentration in the upper box of the model, the maximum ozone values simulated for the lower box of the model increased by 13 to 45 ppb from what they would have been with only 40 ppb of ozone aloft.

For the range of conditions simulated with the box model, roughly 60-75 percent of the maximum ozone seen in the lower layer was due to same-day emissions in the lower layer plus an assumed 40 ppb natural background from aloft; while roughly 25-40 percent was due to carryover and formation in the upper box. However, reducing ozone concentrations in the upper box generally did not decrease the concentrations in the lower box in proportion.

These results imply that reductions in transported ozone might lower the regional background concentrations upwind of OTR urban areas by an amount similar to the reduction in transported ozone, but that these same reductions in transported ozone might not result in proportional decreases in the peak concentrations downwind of the urban areas.

1. EXTENT THAT OTAG MODEL RESULTS REFLECT IMPORTANT TRANSPORT PHENOMENA

The primary objectives of the OTAG modeling were to characterize the contributions to ozone nonattainment in the eastern United States due to regional transport, and to assess the effects of regional emission reductions on peak ozone in identified ìproblem areas.î The OTAG modeling was intended to capture the larger regional-scale features of ozone episodes and to characterize the relative contribution of regional versus local emission sources. It was not intended to accurately simulate the detailed concentration distributions near individual urban areas. For these reasons, the OTAG air quality and meteorological models were applied using a 12 km grid spacing for the Northeast. The meteorological models also had limited vertical resolution. The grid spacing was not adequate to resolve the local variations in emission density or in topography and surface characteristics that are important for estimating the time and location of maximum ozone concentrations near a large urban area, nor was it adequate to resolve fine-scale features of the flow fields.

Due to the nature of the modeling objectives, it was not expected that the OTAG model would accurately simulate many of the fine-scale transport and mixing processes occurring in the Northeast. Detailed analyses to assess how well the model results reflect the transport phenomena described in Section 2 were not performed, but insights can be obtained by examining the formulation of the model as well as data and analyses related to the regional model evaluation tasks for the July 1995 episode. For this episode, the UAM-V air quality model was used with an input wind field generated by the RAMS prognostic meteorological model.

Some specific questions are addressed below.

• How well do model results reflect the differences between surface and aloft transport?

Comparisons of NARSTO-Northeast wind profiler measurements for the 1995 episode with OTAG model wind fields show similar transport characteristics. In most cases, model wind speeds are comparable to measured values at the surface and aloft, recognizing that the model vertical layer structure smoothes out some features of the wind profiles. (See more specific discussions below concerning night jet and coastal effects.) During periods with very light winds, the model surface level wind speeds are often lower than observed. Comparisons between back trajectories derived from the OTAG model and from interpolation of the actual upper-air data (using CALMET) for the July, 1995 episode show similar transport distances over a 24-hour period, for both near-surface and 500 m level winds. However, simulated near-surface wind directions sometimes differ from measured directions by 60 degrees or more. These differences are usually greater for light wind conditions, and at locations where local influences (topography, coastal) are present. For the 500 m elevation, comparisons generally showed qualitatively close agreement between observed and simulated wind directions for July 13 and 14, but differed by up to 45 degrees on the 15th. Such errors in wind direction would lead to errors in the superposition of emissions from different source areas and errors in the mixing of surface emissions with ozone transported aloft.

• How well do model results reflect the night jet and decoupling between surface and aloft flows?

The modeled winds for the 1995 episode show no evidence of a ìnight jetî (e.g., a wind speed maximum a few hundred meters above the surface). The OTAG modeling approach (12 km grid, limited vertical resolution, use of RAMS model) is not able to predict such phenomena. Even if the RAMS were to predict a night jet aloft, use of layer-average winds with layers at 250-500 m and 500-1500 m in the air quality model would wash out much of the effect. However, the model does resolve the decoupling of the aloft flow from the surface, showing light winds at the surface and continued higher-speed flow aloft at night. It also resolves the general clockwise turning of the wind direction with altitude.

It is difficult to isolate the impact of transport associated with the ìnight jetî on air quality. Basically, we know that simulated transport is often incorrect, in terms of where and when a particular air parcel aloft originated, but we do not know how much different the pollutant loading of that air parcel would be if the transport were correctly simulated, or what the effect on peak simulated ozone concentrations might be. In general, however, the model reflects that nighttime ozone transport aloft occurs, even if it does not get the details of the jet correct.

• How well do model results reflect the evolution of the mixing structure?

We did not perform a specific evaluation of the temperature profile/mixing height inputs for OTAG modeling. At elevations below 500 m agl, the RAMS and UAM-V models have adequate resolution to simulate the slow rise of the mixing height on episode days; but no evaluation was performed to see if the mixing height was simulated properly. Above 500 m, the UAM-V vertical layer structure can only account very roughly for mixing height changes since there is only one layer from 500 m to 1500 m. This large layer depth makes it difficult for the model to resolve the effects on ozone transport of the mixing height increasing through the channeled flow regime into the synoptic regime.

• How well do model results reflect coastal flow regimes?

Model input wind fields generally do not reflect the mesoscale influence of the land-sea interface. Review of wind measurements at coastal locations (e.g., Connecticut and Long Island) during the 1995 OTAG episode indicates several days when regional-scale transport was dominant and sea-breeze circulation produced only modest local effects. On July 15, however, failure to account for sea-breeze circulation led to erroneous transport predictions and poor model performance by UAM-V in the Northeast. Measured surface-level wind directions at Brookhaven, NY (on Long Island) and Millstone Point (on the Connecticut shoreline) reflected an onshore WSW sea breeze flow on July 15, while UAM-V model inputs were WNW. This difference was enough for the model to miss the impact of the New York City area plume on the Connecticut shoreline.

The coastal flow regime and local transport around the Chesapeake Bay on July 15 are more complex, and we have less meteorological measurements for that region, but it is also clear that OTAG model results did not correctly predict transport in the B-W subregion on that day.

• How well do trajectories calculated from OTAG model inputs agree with trajectories produced with a diagnostic model using NARSTO-Northeast measurements?

For July 14, trajectories based on near-surface and 500 m level winds show good agreement between the OTAG model and CALMET diagnostic wind fields. RAMS (OTAG) surface-level trajectories were about 25 percent longer (higher wind speeds), but transport directions were similar, ranging from S to WSW. Much closer agreement was seen aloft, with transport from WSW to W.

For July 15, RAMS back trajectories starting at 1500 showed surface-level transport from W to WNW across the entire NE Corridor, with transport at 500 m from WNW to NW. CALMET (observed) back trajectories at inland locations (e.g., Ware, MA and Rutgers) were very similar to RAMS results, both near-surface and aloft. Back trajectories for Millstone Pt., CT, Brookhaven, NY, and Lums Pond, DE, however, all indicated surface-level transport from WSW and 500 m level transport from W for the first 6 hours (back to 0900). (See discussion of coastal influence above.)

From the comparisons of the trajectories and wind fields, it appears that the speeds, directions, and transport distances of the simulated regional flows are qualitatively correct. However, as expected from the coarse model resolution, the near-surface flows are less accurately resolved, resulting in errors in the location of impact of urban emissions.

1. DIFFERENCES BETWEEN MEASURED AND MODELED OZONE CONCENTRATIONS

Lurmann et al. (1997) examined model performance in four subregions within the Northeast Corridor (Baltimore-Washington, Philadelphia, New York, and Boston nonattainment areas) and in the boundary areas immediately upwind of the Northeast Corridor.

The evaluation included (1) a review of observed and simulated ozone concentrations in the Northeast Corridor; (2) comparisons of observed and simulated ozone at ìboundary sitesî located along the western boundary of the Northeast Corridor; and (3) comparison of simulated and observed ozone aloft and transport for the July 1995 ozone episode.

The OTAG air quality and meteorological models were applied using a 12 km grid spacing for the Northeast. This grid spacing is not adequate to resolve the local variations in emission density or in topography and surface characteristics that are important for estimating the time and location of maximum ozone concentrations near a large urban area. The performance analysis was therefore focused on issues pertaining to regional-scale performance, recognizing the practical compromises that were necessary to achieve OTAGís mission with the available data and computational resources.

The following findings emerged from the assessment of the OTAG simulations:

• How well do model simulations agree in time and space to observed surface and aloft ozone and precursor concentrations in the Northeast? Can concentration differences be partially explained by differences between measured and modeled transport phenomena?

Comparisons of observed and simulated ozone concentrations for subregions corresponding to the major metropolitan areas along the Northeast Corridor show more scatter than the regionwide performance statistics reported in the OTAG evaluation. Results for individual subregions and days in the Northeast Corridor include some cases with excellent agreement between the model and the measurements and other cases of poor agreement. The excellent agreement cases have peak ozone levels simulated within 15 ppb of the observed values and within 50 km of the observed peak locations, as well as mean normalized bias and error statistics that meet EPAís acceptance criteria. In contrast, the poor agreement cases have either bias in peak ozone levels of more than ±40 ppb or mean normalized bias and error statistics that fail to meet EPAís acceptance criteria (or both).

The daily maximum concentrations unpaired in space and time compare more favorably with the observed maxima than the simulated maximum concentrations paired in space or paired in space and time. That is, the magnitude of the highest simulated concentration in a subregion is often comparable to the highest observed at any station in the subregion. However, the exact time and location of the simulated daily maximum concentrations are rarely accurate. This mismatch in space and time is partially due to inaccuracies in the modeled wind fields and is typical of air quality model simulations.

On most days, the comparisons of observed and simulated ozone concentrations in the Northeast Corridor show little prediction bias for peak and mean concentrations in ìurban plumes.î The model error is comparable to typical urban-scale modeling results for many subregions and days. However, there are a modest number of days in the 1988 episode and a small number of days in the 1991 and 1995 episodes where the modelís bias and error do not meet EPAís urban-scale photochemical modeling acceptance criteria, suggesting caution about the reliability of the modeling effort. These poor performance days occur in the middle of episodes, as well as at the beginning and end of episodes (see below).

The model performance is less biased for all hourly concentrations above 40 ppb and the daily 90th percentile concentration than for the daily maximum concentrations. The coarse spatial resolution selected for the OTAG simulations (12 km) and differences between the modeled and measured transport winds probably compromise the accuracy of the maximum concentration estimates.

For each episode, the mean normalized bias for ozone shows a systematic trend with time during the episode from underprediction to overprediction (known as bias creep). The model consistently underestimated the observed concentrations for the first few days of the episodes and overestimated the observed concentrations for the last few days of the episodes. This trend clearly demonstrates a regional-scale buildup of simulated ozone that is not evident in the observations. This systematic trend may indicate a problem with the model formulation or model inputs that may influence the accuracy of its simulated response to emissions controls.

Analysis of concentrations along the western boundary of the urban corridor indicates the boundary layer ozone flowing into the corridor was systematically underestimated in the simulations. This bias is evident on most days in all of the episodes. The regional transport contribution to ozone at the boundary of the urban corridor was 5-10 ppb lower than observed on average. The daily maximum concentrations in the upwind area were underestimated by 10-20 ppb on average.

• How do model simulations compare with the observed vertical distributions of ozone and precursors? Are the layers aloft and carryover of ozone resolved by the model?

Comparison of vertical profiles of ozone concentrations (using NARSTO-Northeast aircraft data for 1995) were mixed, with some showing good agreement and others showing poor agreement. The model commonly underestimated the morning ozone aloft and, therefore, the carryover of ozone from the previous day. The model sometimes overestimated the extent of vertical mixing in the morning. The simulated and observed afternoon vertical profiles often showed good agreement and little bias for the Northeast. The coarse vertical resolution in the model simulations limits the modelís ability to resolve the vertical structure, especially in the morning and over the ocean.

Comparisons of simulated and observed vertically integrated ozone concentrations show the UAM-V model simulations underestimate ozone aloft by 10 and 5 ppb on average in the morning and afternoon, respectively. The morning bias in ozone aloft may be important, but the afternoon value may not be significant.

Comparisons of simulated and observed vertically integrated NO_y,, RHC, and CO concentrations show the simulations overestimate NO_y concentrations and underestimate RHC and CO concentrations on average. The vertically integrated concentrations of these species and the biases between model and observations are comparable in the morning and afternoon. The magnitude of RHC underestimation is large. On average, the model estimates 27 ppbC of RHC aloft when 90 ppbC was observed. These results must be interpreted cautiously because the number of data points for this comparison is small. Nevertheless, there is concern because if this RHC bias were widespread, then it would suggest that the VOC emissions are underrepresented in the modeling system and that the model may be getting the right ozone for the wrong reasons.

Overall, the 1991 and 1995 episodes were simulated more accurately than the 1988 episode; this may be largely due to the differences in the meteorological models used in the simulations.

These findings indicate that the estimated ozone transport contributions into the Northeast Corridor and between subregions within the corridor are subject to greater bias and uncertainty than the simulated contributions from nearby urban-scale emission sources (although the location of the urban plumes was not necessarily simulated accurately).

1. REFERENCES

Blumenthal D.L., Lurmann F.W., Kumar N., Ray S.E., Korc M.E., Londergan R., and Moore G. (1997) Transport and mixing phenomena related to ozone exceedances in the Northeast. Prepared for Ozone Transport Assessment Group Air Quality Analysis Workgroup by Sonoma Technology, Inc., Santa Rosa, CA and Earth Tech, Concord, MA, STI-996133-1710-FR, March.

Korc M.E., Roberts P.T., and Blumenthal D.L. (1996) NARSTO-Northeast data management plan. Version 3.1. Report prepared for Electric Power Research Institute, Palo Alto, CA by Sonoma Technology, Inc., Santa Rosa, CA, STI-95141-1537, Research Project EPRI WO9108-01, May.

Lurmann F.W., Kumar N., Londergan R., and Moore G. (1997) Evaluation of the UAM-V model performance in the northeast region for OTAG episodes. Prepared for the Ozone Transport Assessment Group, Air Quality Analysis Workgroup by Sonoma Technology, Inc., Santa Rosa, CA, STI-996133-1716-FR, March.

OTAG (1996a) Modeling protocol. Report prepared by Ozone Transport Assessment Group, U.S. Environmental Protection Agency, Research Triangle Park, NC.

OTAG (1996b) Evaluation of the UAM-V model performance in OTAG simulations: summary of performance against surface observations. Draft report prepared for Ozone Transport Assessment Group, U.S. Environmental Protection Agency, Research Triangle Park, NC, October.

Roberts P.T., Korc M.E., Blumenthal D.L., and Mueller P.K. (1995) Description of the NARSTO-NE 1995 Summer Ozone Study. Version 1.1. Report prepared by Sonoma Technology, Inc., Santa Rosa, CA and Electric Power Research Institute, Palo Alto, CA, November.