Effects of radical source strengths on ozone formation in models for Houston, Texas William Vizuete ∗ , Marianthi-Anna Kioumourtzoglou, Harvey Jeffries, Barron Henderson University of North Carolina Chapel Hill, Department of Environmental Science and Engineering, 113 Rosenau Hall CB#7431, Chapel Hill, NC 27599
Abstract Predicted atmospheric radical source strengths in Houston, Texas, were investigated with regulatory air quality model simulations. The simulation period was August 16 to September 6, 2000, with modeling inputs prepared by the Texas Commission on Environmental Quality. This regulatory model, with these inputs, makes predictions that are often biased high for NO2 and VOC, while underpredicting observed O3 peaks for the episode. Analysis of the base simulation suggested that over much of the Houston area, the model’s radical initiation strengths were more limited for · OH radicals created from primary aldehyde sources, then those from O3 photolysis. These results prompted the creation of a series of “radical source” sensitivity simulation scenarios. These scenarios show that without increased initiation of · OH radicals (or HO·2 radicals), the majority of the VOCs remain unreacted and thus are unable to contribute to ozone formation. While highly reactive VOCs were present in excess, there were an insufficient amount of VOC that actually reacted with · OH. Had these VOCs reacted, they would not only have created more · OH via formaldehyde products, but increased O3 production. In this sensitivity analysis, we show that this model’s ozone production was most significantly increased by adding emitted, or primary sources of formaldehyde. The emitted formaldehyde photolyzed to create HO·2 radicals that became · OH radicals immediately after reacting with NO. Thus, in this regulatory model, the emission and meteorological inputs provided failed to create the environment needed to produce a larger primary source of · OH radicals via either photolysis of carbonyls or nitrous acid. In addition, other potentially large secondary sources of · OH, derived from subsequent reaction products, were lacking. This lack of “new · OH” might be attributable to failure to include sufficient primary carbonyls in the emissions inventory, failure to account in the model for the creation of carbonyls by control devices such as flares, the use of incorrect emissions speciation profiles in industrial or mobile sources, a missing source of nitrous acid, or a combination of all of these. The data suggest that additional ambient observations focusing on the measurements of the magnitudes of potential radical sources are needed. Key words: ozone, process analysis, radical budget, hydroxyl radical, CAMx
Preprint submitted to Elsevier
10 April 2008
1
Introduction
A large region in southeast Texas is in violation of the 8-hour National Ambient Air Quality standard (NAAQS) for ozone (O3 ). The region includes the Houston metropolitan area and the counties of Brazoria, Chambers, Fort Bend, Galveston, Harris, Liberty, Montgomery, and Waller. The region is classified as moderate nonattainment and has been given a maximum attainment date of June 15, 2010. To attain the O3 NAAQS, policy makers must develop a State Implementation Plan (SIP) that contains pollutant reduction strategies. Future attainment can be demonstrated through use of an air quality modeling system that quantifies whether the reduction strategy will bring pollutant levels below the federal standard. The credibility of these reduction strategies is thus contingent on using a modeling system with adequate performance in reproducing atmospheric conditions. A SIP based on a faulty modeling system could lead policy-makers to implement ineffective or counterproductive reduction strategies. The air quality modeling system developed for the Houston SIP is unable to predict the magnitude of the O3 peaks observed in the Houston metropolitan area, despite the fact that O3 precursors of volatile organic compounds (VOCs) and nitrogen oxides (NOx ) are overpredicted. The overprediction of VOCs, specifically ethylene and propylene, are the result of additions made to the base emission inventories by the Texas Commission of Environmental Quality (TCEQ) in an effort to increase VOC reactivity. Even with the large increases, however, these measures have failed to increase ozone production to sufficient levels. The work reported here focuses on another important component of ozone formation, the source and magnitude of free radicals in the model, and estimates the impact that these molecules might have in determining the level of ozone productivity in the Houston airshed. Free radicals are created via initiation processes, which are almost always photolytic reactions. When the radicals react they are either recreated in chain propagation processes such as NO and VOC oxidation, or they are destroyed in termination processes forming stable tropospheric compounds such as nitric acid (Jeffries et al., 1994). The hydroxyl radical (· OH), the most important atmospheric radical, is formed directly by the photolysis of nitrous acid (HONO) or O3 . They can also be formed indirectly by the photolysis of aldehydes and higher carbonyls that produce HO·2 and RO·2 radicals which rapidly become ∗ Corresponding author. Email address:
[email protected] (William Vizuete). URL: http://www.unc.edu/∼vizuete (William Vizuete).
2
· OH
after an NO to NO2 conversion. In this paper, we count the · OH produced from these HO·2 or RO·2 radicals created by aldehyde photolysis as “new” · OH. The · OH radical can react with most trace species and is present in the atmosphere at relatively high concentrations, on the order of 106 molecules/cm3 (Seinfeld et al., 2006). This high concentration is sustained due to a regeneration process that occurs when · OH reacts with VOCs. When hydroxyl radicals oxidize VOCs, they produce hydroperoxy (HO·2 ) or alkylperoxy (RO·2 ) radicals that convert NO to NO2 and recreate the · OH radical. Some of the NO2 then photodissociates, and the atomic oxygen that is created reacts with molecular oxygen to form O3 . In the radical propagation process, the “same” · OH radical (new or re-created) is used multiple times to oxidize VOC, until the · OH reacts in a termination pathway. The measure of · OH radical propagation is quantified by the · OH chain length, or · OH cycle number, which represents the average number of times each new · OH reacts and is recreated before being lost to termination reactions. In a similar manner, the regeneration of NO due to NO2 photolysis can be quantified as a NO cycle. The efficiency of ozone production is determined by the abundance of new radicals, the magnitude of the · OH cycle number, the number of NO to NO2 conversions per · OH cycle, the magnitude of the NO cycle number, and the efficiency of forming O3 through the NO2 photolysis reaction. In this paper, we quantify these terms under a variety of model conditions for Houston, TX. Houston covers a large region, more than 600 square miles (1,600 km2 ), and includes a diverse range of emission sources. In population, Houston is the 4th largest city in the United States, and it produces significant amounts of mobile emissions that are distributed over a large region. The city is also known for its large industrial base, located just east of the downtown area, in the Port of Houston. The Port of Houston, consisting of the Houston Ship Channel and Galveston Bay, is 25 miles long with a mixture of public and private industrial facilities. The facilities located in the Houston Ship Channel form the largest petrochemical complex in the nation, comprising nearly 40% of the nation’s manufacturing capacity. This complex includes more than 400 chemical manufacturing establishments and two of the four largest U.S. refineries. The large population and extensive industrial base of Houston produce a large and staggeringly diverse emission inventory. The complex range of sources translates into large uncertainties in any modeling efforts to simulate the region. This opens the possibility for underrepresented sources of radical precursors. The work presented here shows the results of a series of modeling runs in which emissions of HONO and VOCs in the mobile and industrial sectors of the Houston region were modified. These changes in the emissions are not suggestions of missing emission sources in the inventory; rather, they were made to measure sensitivities in the model to changes in the · OH budget and ozone production. The impact on ozone production due to these modifications was quantified using the air quality model for the Texas SIP.
3
2
Method
Modeling data were generated with the Comprehensive Air Quality Model with extensions (CAMx) version 4.20 (ENVIRON, 2006). CAMx is a U.S. Environmental Protection Agency (EPA) approved Eulerian photochemical grid model that simulates the chemical transformation, transport, and deposition of species in the atmosphere. CAMx was selected for this work because it is used by the State of Texas for attainment demonstrations of the ozone NAAQs. The CAMx model also has the process analysis (PA) option, which was used for the analyses presented here. The TCEQ developed a modeling episode to evaluate its air quality management plans for southeast Texas that spanned August 22–September 6, 2000 (TCEQ, 2006). This period also coincided with an intensive field campaign called the 2000 Texas Air Quality Study (TexAQS), which provided the observational data needed to evaluate model performance (TexAQS, 2000). The modeling domain for the episode, Figure 1, is a nested regional-to-urban scale with grid resolutions of 36, 12, and 4 km. Meteorological inputs required by the model were based on results from the Mesoscale Meteorological Model version 5, known as MM5 (MM5, 2007). The VOC and NOx emission inventories for the modeling episode were prepared by the TCEQ in accordance with EPA guidelines. A MOBILE 6 based inventory was developed for on-road mobile source emissions; emissions for non-road mobile and area sources were developed using emission factors and the EPA’s NONROAD model, using local activity data when available. Biogenic emission inventories were estimated using the global biosphere emissions and interactions system (GloBEIS), with locally developed land cover data (GloBEIS, 2007; Wiedinmyer et al., 2001). Point source emissions were generated based on the TCEQ’s extensive point source data base for regulation and permitting. This database was supplemented by a special inventory survey and additional monitoring data collected in the industrial region of eastern Houston. In an effort to increase ozone production in the SIP air quality model, the TCEQ increased propylene and ethylene emissions by a factor of eight above the base emission inventory (Jeffries et al., 2005). All modeling simulations presented in this work used this modified version of the emission inventory. The analysis presented here focused on one modeling day, August 25, 2000. Measurements made on that day were some of the highest ozone concentrations for the entire episode. Table 1 shows the peak ozone observations for several monitors located in and around downtown Houston. The table also shows the base simulation predicted peaks at the monitor locations. Although the model predicted ozone peaks of up to 159 ppb, the model was unable to reach the levels of ozone seen in 13 of the 14 surface monitors, and at the HROC monitor there was a 63 ppb underprediction. Figure 2 shows the layer one predicted 4
ozone concentrations, with gray arrows denoting the model predicted wind fields. The plots include surface monitor locations, shown in diamond symbols, with observed wind vectors and ozone concentrations. On August 25, winds from the east transported pollutants from the industrial source region across downtown Houston and into residential areas in the north and west, encountering significant highway NOx emissions. The highest ozone concentrations appeared just north of both downtown Houston and the industrial ship channel. Among the days on which winds from the east mixed the industrial and urban parts of the city, August 25 had the highest ozone concentrations, making this day ideal for sensitivity runs that modified industrial and mobile emission sources. The overprediction of ethylene was also evident on this day. Ethylene measurements were made at the Clinton Monitor site located just east of Interstate 610 and south of Interstate 10, labeled C35C in Figure 3. Ethylene measurements were compared with model data (Figure 4) and the model overpredicted ethylene by as much as a factor of 10. The following sections describe the nine sensitivity modeling scenarios performed for August 25 and the methodology used to analyze the results.
2.1
Sensitivity Scenarios
The model’s sensitivity to changes in the radical budget were investigated using nine scenarios labeled A–I in Table 2. In each scenario, a modification was made to the VOC emission inventory that changed the hydroxyl radical (· OH) budget and affected O3 production. No modifications were made to the NOx emissions inventory in scenarios A-F. Modifications to the emission inventory included increases and decreases to the base carbon monoxide (CO) emissions (A,B), increases to the base aromatic emissions (C), increases to the base ethylene emissions (D), increases to the base formaldehyde (HCHO) emissions (E,F), increases to the base nitrous acid (HONO) emissions (G,H), and increases to both base HCHO and HONO emissions (I). In scenario A (Table 2), CO emissions in the model were increased to 400% of base emissions, and in scenario B they were decreased to 25% of base emissions. These modifications were designed to explore the impact of changing the · OH cycle without the introduction of additional · OH radicals. As shown in reactions 1–5, the · OH reaction with CO yielded carbon dioxide and ozone without the production of intermediates. As a result, no additional products, such as aldehydes, were present as a potential source of · OH radicals. The CO competed for the · OH and thus prevented it from participating in termination reactions resulting in an increased · OH cycle. 5
O2 CO + · OH − −→ CO2 + HO·2 HO·2 + NO → NO2 + · OH
(1) (2)
hν
(3) (4)
hν
(5)
NO2 −→ NO + O O + O2 + M → O3 + M
Net: CO + 2O2 −→ CO2 + O3
The remaining scenarios were designed to add · OH radicals either through oxidation products (C,D), or through primary emissions (E,F,G,H,I). In scenario C, emissions of xylene and toluene were increased to 200% of the base emissions. The · OH attack on aromatics resulted in reaction products (aldehydes), which subsequently photolyzed and produced · OH radicals. Xylene and toluene were chosen due to their prevalence in the urban air shed and their high yield of secondary · OH radicals (Jeffries et al., 1994). Production of these secondary radicals, however, still required a source of · OH to initiate the attack on the parent aromatic species. In scenario D, primary emissions of ethylene were increased by an amount that equaled 2% of the base CO emissions. This was done to investigate the possible underprediction of ethylene from mobile sources. Similar to aromatics, the ethylene had to first be attacked by · OH or O3 to produce HCHO, before additional · OH radicals could be produced. Formaldehyde emissions, in the presence of NOx , provide a direct source of · OH radicals as shown in reactions 6 and 7. In sensitivity scenarios E and F, we made different assumptions about possible sources of formaldehyde emissions. hν
HCHO −→ 2HO·2 + CO HO·2 + NO → NO2 + · OH
(6) (7)
One source of formaldehyde could be the flares of petrochemical facilities in the Houston Ship Channel, since incomplete combustion of VOCs in flares could produce significant amounts of formaldehyde. This, however, has yet to be tested in computational models of flare emissions or by direct observations in the field, though there are laboratory measurements that quantify the amount of formaldehyde produced from the combustion of refinery fuel gas. Researchers from the Sandia National Laboratory measured formaldehyde in flue gas, under superstoichiometric conditions, and observed concentrations of ∼1 ppm (Seebold et al., 1997). This suggests that there may be a combustion source of formaldehyde that does not currently exist in the modeling system. Current model predictions of formaldehyde from combustion sources such as flares may also be underestimated because of the assumptions made when developing the emission inventory. For example, current emission model representations of flares do not distinguish among different facility types, such as 6
olefin plants and refineries, which have differing compositions of flare feeds. In addition, the modeling system assumes a uniform destruction efficiency of 99%, without taking into account possible combustion products within the flare. A destruction efficiency of 99% requires ideal conditions, which are not always present in real-world situations. One combustion modeling study found that high momentum flames, which can occur during upset conditions of industrial flares, are very sensitive to crosswinds. In these flares, crosswind speeds above 1 m/s resulted in a sharp decrease in combustion efficiencies, to below 90% (Castineira et al., 2006). This study also found that excess steam in the flare systems significantly decreased flare combustion efficiency, leading to possible releases of large amounts of uncombusted VOCs. Under such conditions, it is conceivable that large emissions of formaldehyde might also occur. To measure the sensitivity of ozone formation to this possible source of formaldehyde, we created sensitivity scenario E. In this scenario, additional HCHO emissions were added to the 12 flares in the model that had the largest VOC emissions. These flares are located east and southeast of downtown Houston, near the area of the Houston Ship Channel shown in Figure 3. An assumption was made that 1% of total VOC being burned in each flare was emitted as HCHO rather than as CO or CO2 . In this scenario, formaldehyde emissions were increased, but no reductions were made to the emissions of CO or CO2 . In large urban areas, motor vehicles could also be a source of formaldehyde emissions, and this emission source could have a significant impact on atmospheric chemistry given its pervasiveness and magnitude. The importance of this has prompted several studies aimed at quantifying the magnitude of these emissions. One in-depth review of several diesel exhaust speciation studies from 1991-2000 concluded that a characteristic diesel exhaust had a formaldehyde to carbon monoxide ratio of 4% (Merritt, 2003). A follow-up study aimed at improving existing speciation measurements for light-duty diesels found the same ratio, to within 0.5% (Fanick, 2005). Additional speciation studies of diesel emissions have reported formaldehyde-to-carbon monoxide emission ratios that ranged from 1-2.7% (Grosjean et al., 2001; Kirchstetter, 1996; Schmitz, 2000; Siegl, 1999; Sigsby, 1987; Zhu, 2003). These studies included international and domestic exhaust VOC characterizations for vehicles manufactured from 1974 to 2000 with multiple fuel blends that included non-oxygenated and oxygenated diesel. Although there is a consensus that formaldehyde-to-CO ratios are higher in diesel than in gasoline vehicles, the literature suggests that formaldehyde emissions from diesel vehicles are highly variable. Therefore, a motor vehicle sensitivity scenario was developed to quantify the impact on ozone chemistry from this possible source of formaldehyde emissions. In this sensitivity run, scenario F, formaldehyde emissions were set to be equal to 4% of the cell’s CO emissions. The photodissociation of nitrous acid (HONO) is another direct source of · OH radicals, and during morning hours, it appears to make a significant contribu7
tion to the · OH budget (Lammel et al., 1988; Li, 1994; Harrison et al., 1996; Neftel et al., 1996). The origin of observed HONO is not fully understood, although it is accepted that one formation pathway is the heterogeneous conversion of NO2 on surfaces (Lammel et al., 1996). Another source could be a formation product in the emissions of motor vehicles. Researchers measuring automotive exhaust in a Caldecott Tunnel study found a ratio of HONO to NOx of ∼0.003, with uncertainty arising from possible heterogeneous reactions on tunnel walls (Kirchstetter, 1996). In another study, simultaneous measurements of HONO and NO2 were made using a differential optical absorption spectroscopy system in the Wuppertal Kiesbegunnel (Kurtenbach et al., 2001). This study reported a HONO to NOx ratio of 8±1 ×10−3 for an average working day with a high traffic density. These data guided the creation of two HONO sensitivity runs labeled scenarios G and H. In these two scenarios, HONO emissions were set equal to 5% and 8% of the cell’s NOx emissions. These percentages were chosen to take into account direct emissions and the possible heterogeneous formation of HONO from emitted NOx . NO emissions in mobile sources were reduced by the molar equivalent of HONO production. Finally, scenario I combined scenarios E, F, and G, taking into account direct production of · OH from HCHO from flares, and HCHO and HONO from mobile sources. Instead of adding formaldehyde to 12 flares, however, every flare that emitted VOCs in the modeling domain had formaldehyde emissions increased, following the same methodology used in scenario E. The amount of formaldehyde that was added as a possible mobile source was reduced from 4% to 1% of CO emissions. The levels of HONO were also reduced to 0.8% of NOx emissions.
2.2
Process Analysis Method
The CAMx Process Analysis (PA) extension integrates the chemical and physical process rates occurring in the model and writes these values to two output files called the Integrated Process Rate (IPR) and Integrated Reaction Rate (IRR) files (ENVIRON, 2006). The IPR file contains physical process data for every species at each grid cell, layer, and output time step. The process rates include advection, diffusion, deposition, and net chemistry. The IRR file contains data detailing the mass throughput of every reaction in the model’s chemical mechanism. This information can be used to determine how net changes in pollutant concentrations are dependent on individual reaction pathways or classes of pathways. Data from these files are extracted using a program that aggregates each process rate over various areas and mixing depths. Thus, all rates are aggregated within a user specified collection of grid cells called an analysis volume. The concentrations and magnitudes of process rates are an average over the entire analysis volume. Transport rates among 8
the cells and layers within the analysis volume are summed out, and only transport through the analysis volume’s faces remains. This is reported as a single total net exchange at the faces. In this way, the impact of the chemical rates is enhanced while the impact of transport is mimimized. The analysis volume used for this study is shown in Figure 3; it had an area of 1,250 km2 (6×13 4-km grid cells) and encompassed the most photochemically active region in the modeling domain. The area’s size was specifically chosen to include both the Houston Ship Channel and downtown Houston and thus represents an average condition of this main O3 producing region. Therefore, our results were not influenced by “hot emission” cells or other non-typical inputs that might be present in the model. Vertically the analysis volume followed the time-varying planetary boundary layer for each grid cell, minimizing vertical transport out of the analysis volume. The results reported here focus on hours 7-17, which correspond to the daylight period of August 25, 2000. All PA results shown are summed across these daylight hours.
3
3.1
Results
Ozone Results
Table 1 shows the peak ozone observations on August 25 for every monitor located in the analysis volume (Fig. 3). The order of monitors in the first column is based on their location in the analysis volume, from the most western monitor, BAYP, to the most eastern monitor, HO7H. The fourth column of the table shows the base simulation peak ozone predictions at each monitor location. Columns 5–13 show the changes in O3 from the base prediction that resulted in each sensitivity simulation. The addition of carbon monoxide raised peak ozone concentrations, though mainly in the western monitors, and resulted in 11 underpredictions and 3 overpredictions of measured ozone peaks. Similar results also occurred when ethylene and aromatics were added to the base simulation. This suggests that these three species have the greatest effect on ozone concentrations in aged air masses. The aromatic addition not only had an impact on ozone concentrations in the western domain, but also caused a 12 ppb increase in the downtown Houston monitors of HCFA and HROC. Even with this increase, there were still 11 monitors with observed ozone peaks that could not be predicted. Among scenarios A–H, the two with formaldehyde additions, scenarios E and F, had the highest increases in peak ozone concentrations, up to 34 ppb. In scenario E, the formaldehyde additions were spread throughout the domain, which resulted in peak ozone concentration increases in excess of 10 ppb at nearly every monitor in the analysis volume. This widespread increase in ozone resulted in fewer underpredicted 9
monitors, a drop from 13 to 7. The largest O3 changes were predicted at monitors closest to the interstate highway systems, such as HROC and C35C. In scenario F, formaldehyde was added only to flares, limiting the spatial impact on ozone production. Monitors that were downwind of an imputed flare predicted the largest increases in peak ozone. For example, the C35C monitor was predicted to have an increase of 33 ppb. The impact on ozone production dropped farther away from the imputed sources, but was still significant, as shown at the WILT monitor, where the model predicted a 21 ppb increase. The added formaldehyde not only produced higher ozone concentrations, but caused the ozone peak to occur at an earlier time. For example, in the two formaldehyde scenarios, the hour of peak ozone concentration was an hour earlier at the C35C and DRPK monitors. In the simulation of scenario I, with a combination of increases in nitrous acid and formaldehyde emissions, the model predicted the greatest increase in peak ozone and reduced the number of monitors that were underpredicted from 13 to 6. The time of the peak ozone concentration also shifted 1–2 hours earlier for the C35C, HO3H, HO4H, and DRPK monitors.
3.2
Process Analysis Results
Table 3 shows the process analysis results for all simulations. Each row describes a key chemical parameter that is calculated by the process analysis program. Chemical parameters have been described in detail in Jeffries et al. (1994). Ozone chemical production rates were most sensitive to scenarios with increased HCHO emissions (Scenarios E and F), or the combination of increased HCHO and HONO emissions (Scenario I). Line 18 shows the total amount of O3 that was chemically produced in each scenario, and line 20 reports the maximum O3 concentration for each scenario averaged across the entire analysis volume. The simulation of scenario I predicted an increase in the analysis volume averaged ozone concentration by 33%, and the simulations of scenarios E and F resulted in increases of 12 ppb. The changes to the model’s emission inventory in scenarios E, F, and I affected not only the magnitude of O3 concentrations, but also the places where these changes occurred. Figures 6 and 7 are plots of layer one predicted O3 differences, for hours 13–16, in sensitivity scenarios D and E. In scenario D (Figure 6), ethylene was added with existing CO emissions, resulting in additions mainly along highway systems and in downtown Houston. The additional ethylene increased O3 concentrations in aged air masses where elevated concentrations already existed. This is the reason for the minimal increases in peak ozone concentrations reported for the eastern monitors (Table 1). The simulations of scenarios A–C had similar results. In scenario E, formaldehyde emissions were added to 12 flares with the largest VOC emissions (see Fig. 3, and note that three of these flares were outside the analysis region to the east). In contrast to the simulation where 10
ethylene was added, the formaldehyde in this simulation had an immediate impact on O3 concentrations near the location of the emission source (Figure 7). The peak ozone concentrations at all 14 monitors listed in Table 1 increased by as much as 34 ppb. Table 3 quantifies several of the radical processes that are important to ozone production. The chemical process chain that leads to O3 production begins with radical initiation, or the production of new · OH. The total amount of new · OH created (line 7), multiplied by the · OH chain length (line 8), is the total amount of · OH that can react with CO, VOCs, and NO2 (lines 10-12). In these scenarios, the · OH chain length was essentially constant; thus differences in the amount of total · OH that reacted were dependent on the magnitude of new · OH sources. Sources of new · OH radicals include the photolysis of aldehydes and HONO, and the creation of · OH via O3 photolysis. Since · OH radicals are required to make O3 , the creation of · OH via O3 photolysis represents a positive feedback strongly dependent on the magnitude of · OH. A modeling system that has a limited amount of · OH translates to less O3 production, dampening this positive feedback. In previous modeling studies of O3 chemistry, sources of new · OH radicals were evenly distributed between inorganic (HONO,O3 ) and organic (aldehydes) precursors (Jang et al., 1995a,b; Jeffries et al., 1994). As shown in Table 3, the organic sources of · OH made up a smaller fraction in these model simulations, approximately ∼43%. The first line in Table 3 reports the total amount of HCHO that photolyzed, producing two new HO·2 radicals that then became two new · OH radicals at an efficiency greater than 90%. As expected the CO, aromatic, and HONO additions had little effect on total HCHO photolysis. Line 2 gives the amount of higher aldehydes and other organic VOCs that photolyzed. This value remained unchanged for all sensitivity simulations except the one with the addition of aromatics (scenario C), which more than doubled the base simulation value. Lines 3 and 4 report the amount of ozone and nitrous acid that photolyzed. As expected, the simulation that produced the largest amount of ozone, scenario I, also photolyzed the greatest amount of ozone. In nearly every simulation, nitrous acid contributed a negligible amount to the total new · OH. The addition of nitrous acid in scenarios G and H, however, did increase · OH production from this pathway by three orders of magnitude. Lines 5 and 6 report the amount of new · OH from both organic and inorganic sources, and line 7 is the sum. The model simulations of the CO scenarios had a minimal effect on the total new · OH, increasing it by only ∼3%. This increase was due soley to increased O3 photolysis produced from the higher O3 concentrations. The addition of CO resulted in a shift from · OH termination reactions to reactions with CO (line 10) and produced the largest increase in the · OH chain length (13%). This did not, however, increase the amount of · OH that reacted with other VOCs, as shown in line 11. That means that the increase in new · OH was a feedback process that resulted solely from the photolysis of chemically 11
produced O3 . The simulations of scenarios C–F, and I produced the largest increases in total new · OH, mainly from organically derived new · OH. With the aromatic addition, increases mostly came via higher aldehydic products, as shown in line 2. The emissions change in each sensitivity scenario affected not only the magnitude of · OH concentrations, but also the time when it occurred. In scenarios C and D, increases in · OH concentrations occurred much later in the morning, at hours 9 and 10. In contrast, when HONO or HCHO was added, increases began in hours 6 and 7. The two-hour difference is due to the necessity of an · OH attack for the aromatic or ethylene species to produce · OH radicals. The HCHO and HONO scenarios produced radicals immediately and did not require an initial · OH attack. The largest increases in · OH radicals for scenarios C and D occurred in the western side of downtown Houston. The largest increase was ∼0.38 ppt in scenario I at hour 13 just southwest of downtown Houston. Process analysis data also outputs the amount of · OH that is produced and reacted hourly. The largest increases in the amount of · OH that reacted occurred in the morning hours for scenarios B–I, and began to diminish by hour 14. This was due in part to the reduced production of new · OH radicals in the afternoon hours, but also to a decrease in the · OH cycle number. The decrease in the cycle number in the afternoon was caused by the reduced afternoon levels of NO, decreasing the efficiency of converting HO·2 to · OH via NO to NO2 oxidation. Increases in the magnitude of total new · OH radicals are significant because of the · OH chain length (line 8), or the recycling of · OH that occurs in VOC oxidation. The product of lines 7 and 8 gives the total amount of · OH that could react and lines 10–12 show how this · OH reacted. Radical propagation oxidized CO and VOCs (lines 10 and 11), producing HO·2 and RO·2 radicals. These species mostly reacted with NO to produce NO2 (line 17). Photolysis of the NO2 then led to O3 production, shown in line 18. Radical termination was dominated by the · OH reaction with NO2 , which led to nitric acid (line 12), with similar results in each scenario. This was the result of the ample NOx available in central Houston (line 15). Increased · OH production translated into a greater amount of · OH reacting with VOC; the largest increase was 13 ppb in Scenario I. In each scenario, the amount of VOC that reacted was limited not by the availability of VOCs, but by the competition for · OH radicals by VOCs, CO, and most importantly, NO2 . In most scenarios, approximately one-third of the available VOC reacted with · OH; the rest was transported out of the analysis volume. Although more than 200 ppb of VOC were available for reaction in the base scenario (line 13), only 66 ppb or 32% of the VOC reacted with · OH over the entire daylight period (line 14). This percentage of reaction increased only slightly to 36% in scenarios E and F, and to 39% in scenario I. This reaction is a necessary condition for the production of HO·2 12
and RO·2 radicals that oxidize NO and ultimately lead to O3 production. The average number of these peroxy radicals produced per · OH reaction with VOC was 2.1. As a result, the 66 ppb of VOC that reacted in the base simulation produced approximately 140 ppb of NO2 . In scenario C, additional · OH radicals that formed from secondary carbonyls increased the amount of VOC reacted. Because of the delayed production, however, less O3 was produced than in the HCHO scenarios. A similar result occurred for the ethylene scenario. Large gains in the NO oxidized and O3 produced occurred when direct sources of organically derived new · OH were introduced via the HCHO addition scenarios. For example, the simulation of scenario F produced a 41 ppb increase in peak O3 and increased the analysis volume averaged peak O3 by nearly 12 ppb. Figure 8 shows the distribution of VOCs, CO, and NO2 that the · OH radical reacted with in the analysis volume in the base simulation and in each sensitivity scenario. In the base simulation, 24% of the · OH reacted with NO2 , with similar results in all sensitivity scenarios. When CO emissions were changed (scenarios A and B), the largest changes were seen in the amount of · OH reacting with CO. Increases in · OH reacting with aromatics and ethylene (HRVOCs) were not as dramatic. The HONO addition (scenarios G and H) made very little change in the · OH reaction distribution. The addition of formaldehyde in scenarios E, F, and I not only increased the total magnitude of · OH reacted, but also increased the amount of · OH that reacted with formaldehyde. With process analysis data, we can distinguish this formaldehyde between emitted (primary), and HCHO that was created via oxidation of VOCs (secondary). In the base simulation, a small fraction of the formaldehyde that reacted was primary. This was also true in every sensitivity scenario that did not include an increase in formaldehyde emissions. When formaldehyde emissions were increased (scenarios E, F, and I) a large fraction, up to 50%, of the formaldehyde that reacted with · OH was primary.
3.3
Ambient Measurement Comparison
Modeling predictions from all scenarios were compared with observations of ambient · OH concentrations during the 2000 Texas Air Quality Study (TexAQS) field campaign. During the campaign, researchers from Pennsylvania State University measured · OH concentrations at the La Porte monitor (LAPT) with the GTHOS instrument (Ground-based Troposhperic Hydrogen Oxides sensor) by laser induced fluorescence of · OH (Martinez et al., 2002). This monitor is located at the La Porte Airport, adjacent to the Houston Ship Channel and is denoted by the green box labeled LAPT in Figure 3. Figure 9 shows a time series for the model predicted and observed · OH concentrations. At this location, the base simulation underpredicted · OH concentrations during the morning hours. Unfortunately, observational data were available for only 3 13
hours in the afternoon. At hours 12 and 16, the base simulation showed good agreement, but it underpredicted concentrations at hour 13. In the evening hours, a sustained · OH concentration of ∼0.1 ppt was observed; model predictions were nearly zero and remained unchanged in all modeling scenarios. In all sensitivity studies, except for scenario B, there were increases in · OH concentrations at every daytime hour. In scenario B, a combination of dampened · OH cycles and reduced inorganic · OH sources led to lower predicted · OH concentrations. The addition of aromatics and ethylene caused increases in · OH concentrations by up to 10%, while the addition of formaldehyde and nitrous acid resulted in increases of up to 15%. Scenario I produced the largest increase in · OH concentrations by nearly 75% in the morning hours, as shown in Figure 9. This was due to the fact that the LAPT monitor was sited just east of downtown Houston near the industrial sector where the imputed flares were located. Since the effects of the formaldehyde and HONO additions are immediate, they will register in this part of the modeling domain. In the remaining scenarios, the largest increases in · OH concentrations occurred further downwind, west of this monitor.
Model predictions were compared with formaldehyde concentrations observed on August 25 from three surface monitors labeled the Houston Regional Monitoring Network 3 site (HO3H, EPA site 48-201-0803), La Porte (LAPT), and Williams Tower (WILT). All monitor locations are shown in Figure 3. The HO3H site, located at 1504 21 Haden Rd., collected measurements using an automated fluorometric approach, based on the Hantzsch reaction, with a Nafion membrane DS collector (Li et al., 2005). The HO3H site is located east of downtown Houston in the midst of the petrochemical and chemical manufacturing complex. The HO3H site is operated by the Texas Commission on Environmental Quality; the data from this site are web accessible (TCEQ, 2007). The measurements at the LAPT monitor were made with a differential optical absorption spectroscopy (DOAS) instrument (Harder et al., 1997; Perner et al., 1980; Stutz et al., 1997, 2004; Wert et al., 2003). Measurements were also taken at the 62nd floor (243 m above ground level) of the Williams Tower on the west side of downtown Houston, using a continuous derivatization/fluorescence monitor (Berkowitz et al., 2004; Kelly et al., 1994). On August 25, the base simulation predicted concentrations that were biased low for surface measurements made near the ship channel, but showed good agreement downtown at the Williams Tower (Figure 10). The sensitivity scenarios made significant changes at the surface monitors, especially at night. Figure 10 shows the changes in formaldehyde due to imputations made in scenario I. At the HO3H monitor, the model showed overpredictions for most of the day, with underpredictions at hours 13-15. Model predictions at the LAPT monitor also showed large nighttime overpredictions, but reasonable agreement during the day, and captured the morning peak. There was little change made to predictions at the Williams Tower. 14
In addition to surface observations, measurements were taken aboard the National Center for Atmospheric Research L-188C Electra aircraft using a tunable diode laser absorption spectrometer (Fried et al., 1998, 2002; Wert et al., 2002, 2003). Formaldehyde measurements collected aloft were compared with predicted concentrations from each modeling scenario. Figure 11 shows an aircraft transect made by the NCAR aircraft on August 25 at hours 13–14. The plots show measured formaldehyde and predicted concentrations from the base simulation. When the base model underpredicted ozone concentrations, it also did not reach the measured levels of formaldehyde. The sensitivity scenarios had little impact on concentrations at aircraft altitudes. The largest changes to predicted formaldehyde aloft occurred in scenario I, where a 2 ppb increase in formaldehyde concentrations resulted in a 3 ppb increase in ozone. This was still insufficient to match the observed levels of ozone or formaldehyde. The surface and aircraft comparisons suggest that over large areas, the model is unable to predict observed formaldehyde levels.
4
Conclusions
The modeling results from this sensitivity analysis demonstrate that O3 production is sensitive to changes in organically derived · OH radicals. Formaldehyde additions, a direct photolytic source of · OH radicals, resulted in the greatest increase in O3 production (33% peak increase). Only with formaldehyde was there an immediate impact on ozone production near the release point and in the NOx rich downtown Houston area. Process analysis of the modeling results showed that the major determining factor in O3 production was the magnitude of new · OH radicals. The largest increases in new · OH radicals were produced from the photolysis of aldehydes. The emission inventory used in all simulations included an artificial addition of ethylene and propylene by the TCEQ (TCEQ, 2006). This addition resulted in the base simulation predicting two to five times the measured ethylene and propylene at several monitor locations. Although there were ample VOCs available, because of the lack of new · OH radicals, nearly two-thirds of the available VOCs remain unreacted and thus were unable to contribute to ozone formation. These VOCs were merely advected out of the modeling domain and were unable to contribute enough reactivity to permit the model to predict observed O3 peaks. These results stongly suggest that the low ozone productivity in the Houston SIP model is attributable to a lack of new radicals, rather than to internal radicals generated from reactions with VOCs. All modeling simulations were compared with surface measurements of · OH concentrations and both surface and aircraft measurements of formaldehyde. In all sensitivity studies, with the exception of scenario B, there were increases in · OH concentrations at every daytime hour; nighttime concentrations in all 15
scenarios were predicted near 0. In scenario B, the combination of dampened · OH cycles and reduced inorganic · OH production resulted in a decrease in predicted · OH concentrations. The addition of aromatics and ethylene increased · OH concentrations up to 10%, while the addition of formaldehyde and nitrous acid produced increases of up to 15%. Scenario I saw the largest increases of · OH concentrations, up to 75%, which occurred in the morning hours. Predicted formaldehyde concentrations in each sensitivity scenario were compared with both surface and aircraft data. With the exception of the Williams Tower monitor, the sensitivity scenarios significantly increased surface concentrations of formaldehyde at night. In contrast, when compared with daytime measurements, there were only small increases in predicted formaldehyde concentrations. The base model consistently underpredicted formaldehyde aloft in the aircraft measurements, especially when encountering a high ozone plume. The sensitivity scenarios had little impact on concentrations at the aircraft altitudes. The largest changes to predicted formaldehyde aloft occurred in scenario I, where a 2 ppb increase in formaldehyde concentrations resulted in a 3 ppb increase in ozone. The ozone production in the modeling system is capable of responding to changes in the radical budget. It is the emission and meteorological inputs provided to the modeling system that fail to create the environment needed to photolyze carbonyls or nitrous acid to produce additional · OH radicals. This could be attributed to either a failure to include sufficient primary carbonyls in the emissions inventory or a failure to take into account the creation of carbonyls by control devices such as flares. Other possibilities could be the use of incorrect emissions speciation profiles in industrial or mobile sources, or a missing source of nitrous acid. Future work should include characterizing emission sources. Ambient HCHO and HONO observations are also needed to spatially and temporally characterize this important radical source and to confirm the quality of the emission inputs to the models.
16
Fig. 1. The horizontal and vertical modeling domain used in the CAMx simulations described in this study. The East US (8-Hour), Regional (1-Hour, MCR), East Texas, Houston Galveston Beaumont Port Arthur (HGBPA), Houston Galveston (HG), and Beaumont-Port Arthur (BPA) have nested domains of 36, 12, 4, and 1 km resolutions. The East US (8-Hour) 36 km modeling domain, the 12 km East Texas domain, and the 4 km HGBPA domain were used in the simulations presented in this paper. The BPA and HGB 1 km resolution domains and the 36 km Regional domain were not used. The vertical layers shown in this figure are for the 4 km resolution grid (TCEQ, 2006).
17
Fig. 2. The base simulation predicted ozone concentrations in layer one for hours 13-16 on August 25, 2000. The plots include surface monitor locations, shown in diamond symbols, with observed wind vectors and ozone concentrations. Gray arrows represent model wind vectors and show a westerly to northwesterly flow field. The ozone peak is located just north of downtown Houston at hour 13 and then travels northwest until it leaves the domain at hour 16.
18
436
# 440 *
444
448
452
456
460
464
468
472 # *
476
480
484
488
90 t u 59 t u
) !
-1096
261
) ! -1100
# * # *
HLAA
) !
t u
) !
HWAA
# *
t u
" $ #
90 Alt # * # * # *
) !
-1108
) !
# *
) !
H03H
*# )# ! * # *
# *# *# *# * # * * # # * # *# * # * # *# * # * # *# *# # * # * # * # * # # * * * ( ! * * * # # # *# # *# * 225 (# ! ( * ! *# # *# *
HROC
) !
C35C
) !
59 t u
t u
90 Alt
# *
BAYP
-1116 # *
610 " $ #
-1100
) !
330
# *
H07H
) !
) !
201
H11H
# *
-1104
# *
* ## *
-1108
) !
DRPK
) !
* )# ! # *# *! (
# *
*# ## * * * # *# * # * # # * * # (# ! (* ! (# ## *! * # * ** * 134 # *# # * # * # * # * # * # # * # * * # # * *#
) !
) !
146
* ## *
-1112 # *
LAPT
-1116
H08H
# *
35 ) !
-1120
) !
HOEA
# *
HCFA
WILT
) !
# *
) !
HCQA
# * # *
HSMA
# * # *
" $ # 45
* )# # *! # *# * # * # * # *# * *# * # *#
-1120 # * # *
# * # *
)3 !
-1124
-1096
# *
10
# *
# *
10 " $ #
-1104
) !
* # *#
# * # * # *
290
-1112
) !
H04H
# *
# * 492 # * ( ! # * # * # #* * * # *# (# ! * # *# * 146
-1124
# * # * * #
# *
-1128
-1128 # *
# * # *
-1132
-1132 # *
-1136
# *
# *
) !
-1140
-1136
) !
146
)6 ! ) !
35 Bus
# *
288
-1140 # *
-1144
) !
Monitors
( !
Imputed Flares
# *
) !
440
-1144
( ! # *
Point Sources 436
TLMC
444
448
452
456
460
464
468
472
476
480
484
488
* *# # *# # * # * * # # * * # *# *# 492
Fig. 3. Map of downtown Houston and the Houston Ship Channel, with light gray lines showing the location of the model’s 4×4 km grid cells. The central Houston analysis volume is outlined in red. The analysis volume has an area of 1,250 km2 (6×13 4 km grid cells) incorporating both the Houston Ship Channel and downtown Houston. Green squares represent ground monitor sites, triangles the location of industrial point sources, and red circles the location of the imputed flares used in scenario E (addition of 1.3 t/d of fromaldehyde).
19
Fig. 4. Observed ethylene concentrations measured at the Clinton monitor (C35C) versus model predictions from the base simulation on August 25, 2000. Increased ethylene emissions, beyond the base emission inventory, resulted in an overprediction of ethylene by as much as a factor of 10.
20
Fig. 5. Model predicted layer one · OH concentration differences for hours 10-13 on August 25, 2000. Differences were calculated by using predicted concentrations in scenario I (addition of 1 t/d of formaldehyde and 0.2 t/d of nitrous acid) minus the base scenario. The addition of formaldehyde and nitrous acid in this scenario resulted in an increase in · OH concentrations near the emission sources.
21
Fig. 6. Predicted ozone concentration differences in layer one for hours 13-16 on August 25, 2000. The values were calculated using predicted ozone concentrations from scenario D (addition of 2.8 t/d of ethylene) minus predictions from the base simulation. The largest increases were seen in the western part of the domain, where the largest ozone concentrations were predicted. These enhancements followed the ozone plume as it traveled northwest out of the modeling domain.
22
Fig. 7. Predicted ozone concentration differences in layer one for hours 13-16 on August 25, 2000. The values were calculated using predicted ozone concentrations from scenario E (addition of 1.3 t/d of formaldehyde from flares) minus predictions from the base simulation. The largest increases were seen immediately at the imputed flares, where formaldehyde emissions were increased. These enhancements followed the ozone plume as it traveled northwest out of the modeling domain.
23
140
120
Total OH Reacted (ppb)
secondary FORM primary FORM
100
secondary ALD2 primary ALD2 HRVOC
80
biogenic VOC aromatic VOC 60
ETOH/MEOH PAR CH4
40
CO NO2
20
I ar
io
H io
io
ar
Sc en
G Sc en
ar
io
Sc en
ar
io
F
E Sc en
ar
io
io
D Sc en
ar Sc en
ar
io
C
B Sc en
ar
io
A Sc en
ar Sc en
Ba
se
0
Fig. 8. The distribution of VOC species that the · OH radical reacted with in the analysis volume for the base simulation (first column) and all sensitivity scenarios. Data were aggregated across hours 7-17 on August 25, 2000. The simulation of scenario I (addition of 1 t/d of formaldehyde and 0.2 t/d of nitrous acid) had the largest increase in total · OH reacted, and also the largest increase in the amount of · OH that reacted with primary (emitted) formaldehyde.
24
Measured
Base
Scenario I
Fig. 9. Ambient · OH concentrations made at the Laporte surface monitor and model predicted · OH concentrations in the base simulation and for scenario I (addition of 1 t/d of formaldehyde and 0.2 t/d of nitrous acid). The base simulation underpredicted · OH concentrations during the morning and evening hours. The simulation of scenario I produced the largest increase in · OH concentrations, up to 75%, which occurred in the morning hours
25
HO3H
WILT
Measured
LAPT
Base
Scenario I
Fig. 10. Observed formaldehyde concentrations from the WILT, HO3H, and LAPT surface monitors. Also included are the model predicted formaldehyde concentrations from the base simulation and scenario I simulation (addition of 1 t/d of formaldehyde and 0.2 t/d of nitrous acid).
Fig. 11. Measurements taken aboard the National Center for Atmospheric Research L-188C Electra aircraft on August 25, 2000, at hour 13-14 LST while the aircraft was flying transects across Houston. Also shown are the base simulation model results, labeled b1b_psito2n2. When the base model underpredicted ozone concentrations, it also did not reach the measured levels of formaldehyde. Also note the large overprediction of CO in the model.
26
Table 1 Observed and modeled peak ozone concentrations (ppb) on August 25, for every monitor located in the analysis volume shown in Fig. 3. The order of monitors in the first column is based on their location in the analysis volume from the most western monitor, BAYP, to the most eastern monitor, HO7H. The fourth column of the table shows the base simulation peak ozone predictions at each monitor location. Columns 5–13 show the change to that ozone prediction that resulted from emissions changes in each sensitivity scenario (sensitivity scenario minus base simulation). Changes in peak ozone (ppb) resulting from emission modifications Carbon Monoxide Monitors
Aromatics
Ethylene
Formaldehyde
Nitrous Acid
Radical Sources
Hour
Observation
Base
A
B
C
D
E
F
G
H
I
BAYP
14
138
130
14
-5
15
10
15
20
7
11
30
HLAA
15
177
159
10
-4
9
6
18
11
4
6
28
WILT
14
172
142
12
-5
16
9
27
21
8
12
47
HCFA
14
194
134
8
-3
12
6
27
13
5
8
43
HROC
13
185
122
7
-3
12
5
34
14
6
9
58
HWAA
13
155
150
6
-3
5
4
14
8
3
4
22
C35C
13
157
103
5
-2
10
4
33
14
4
8
61
HOEA
12
155
110
5
-2
9
4
31
13
5
9
53
H03H
12
119
92
4
-2
7
3
29
9
4
6
51
H04H
13
79
104
3
-1
1
1
6
2
0
1
20
DRPK
13
132
81
3
-2
3
2
14
5
1
2
30
LAPT
12
87
73
3
-2
2
2
9
3
1
2
7
H08H
12
79
76
2
-2
2
1
5
2
1
1
5
H07H
12
130
85
2
-1
1
1
6
2
1
1
8
Model = Obs.
0
0
0
1
0
1
0
0
0
0
Model < Obs.
13
11
13
11
12
7
11
13
12
6
Model > Obs.
1
3
1
2
2
6
3
1
2
8
27
Table 2 Sensitivity scenarios used in this study listing target VOC species and changes to emission rates in tons per day. All emission rate changes were made with reference to the base simulation and limited to the HGBPA 4 km modeling domain shown in Fig. 1.
Scenario Species A carbon monoxide
Total (tons/day) 505
B
carbon monoxide
C
xylene toluene
1.0 0.5
D
ethylene
2.8
E
formaldehyde
1.3
F
formaldehyde
5.6
G
nitrous acid nitric oxide
0.9 -0.9
H
nitrous acid nitric oxide
1.5 -1.5
I
formaldehyde nitrous acid nitric oxide
1.0 0.2 -0.2
28
-126
Table 3 Process analysis chemical parameters in the analysis volume for the base simulation and the nine sensitivity scenarios on August 25, 2000, summed over hours 7-17. All values are in ppb unless denoted by a percentage. The analysis volume’s horizontal domain includes 6×13 4 km grid cells vertically aggregated for the time-varying planetary boundary layer depth. Resulting process analysis paramaters from emission modifications Carbon Monoxide Parameter (ppb)
Base
Aromatics
Ethylene
Formaldehyde
Nitrous Acid
Radical Sources
A
B
C
D
E
F
G
H
I
HCHO+hv
5.6
5.5
5.6
5.7
6.1
6.9
7.6
5.7
5.8
8.4
(ALD2a, otherb)+hv
1.7
1.8
1.7
2.6
1.8
1.8
1.8
1.8
1.8
1.8
O3 +hv
8.8
8.6
9.3
9.3
9.2
9.9
9.8
9.2
9.4
10.7
0.005
0.005
0.005
0.005
0.005
0.005
0.005
1.8
2.6
0.029
organic new OH
14.4
14.4
14.4
16.0
15.4
16.8
18.4
14.8
14.9
19.5
inorganic new .OH
19.6
20.6
19.0
20.6
20.3
21.7
21.6
21.4
22.8
23.7
total new .OH
34.0
35.0
33.4
36.6
35.7
38.6
40.0
36.2
37.7
43.2
.
3.2
3.6
3.0
3.2
3.2
3.2
3.2
3.2
3.2
3.2
.
108
126
102
118
116
123
129
116
120
140
.
OH+CO
14.4
32.5
8.8
15.7
15.1
16.4
17.8
15.9
16.8
18.8
.
OH+VOC
65.8
65.3
65.6
73.6
72.1
77.0
79.8
70.2
72.6
88.2
.
24.0
24.2
23.5
24.9
24.7
25.7
27.4
25.6
26.4
27.6
HONO+hv .
OH chain length OH reacted
OH+NO2
VOC avail.
206
206
205
210
215
214
223
206
205
227
% VOC Reacted
32%
32%
32%
35%
34%
36%
36%
34%
35%
39%
NOx avail.
75.8
75.8
75.9
75.8
75.8
75.6
75.6
73.6
72.4
74.9
2.7
2.7
2.7
2.8
2.8
2.8
3.0
2.8
2.8
3.0
NO+(HO ,RO )
148
163
142
163
159
166
174
156
161
184
O3 produced
105
120
100
119
115
123
129
113
118
140
0.7
0.7
0.7
0.7
0.7
0.7
0.7
0.7
0.7
0.8
O3 peak
87.1
93.3
84.6
94.1
91.7
99.0
98.9
91.5
94.1
108.9
O3 Horz. export
73.2
84.2
69.5
83.2
81.1
83.3
90.1
79.2
82.6
94.1
VOC avail./NOx avail. . 2
. 2
. 2
. 2
O3 produced/NO+(HO ,RO )
a
ALD2 = Carbon Bond IV representation for higher aldehydes b other = Carbon Bond IV species OPEN, MGLY, and ISPD OPEN: aromatic ring opening product MGLY: methylglyoxal & other aromatic product ISPD: isoprene product (lumped methacrolein, methyl vinyl ketone, etc.)
29
References Berkowitz C. M., Jobson T., Jiang G., Spicer C. W., Doskey P. V., “Chemical and meteorological characteristics associated with rapid increases of O3 in Houston, Texas” J. Geophys. Res., 109, D10307, doi: 10.1029/2003JD004141, 2004. Castineira D., Edgar T.F., “CFD for simulation of steam- assisted and airassisted flare combustion systems” Energy & Fuels, 20, 1044-1056. ENVIRON Holdings, Inc. 2006, User’s Guide: CAMx Comprehensive Air Quality Model with Extensions Version 4.20, Available at http://www. camx.com Fanick Robert E.,Diesel Exhaust Standard - Phase II: CRC Project No. AVFL10b. Final report, Southwest Research Institute, August 2005. Fried A., Henry B., Wert B., Sewell S., Drummond J. R., “Laboratory, groundbased, and airborne tunable diode laser systems: Performance characteristics and applications in atmospheric studies” Appl. Phys. B, 67, 317 330, 1998. Fried A., Lee Y.-N., Frost G., Wert B., Henry B., Hubler G., Jobson T., “Airborne CH2O measurements over the North Atlantic during the 1997 NARE campaign: Instrument comparisons and distributions” J. Geophys. Res., 107, 4039, 10.1029/2000JD000260, 2002. Global biosphere emissions and interactions system (GloBEIS) 2007, http: //www.globeis.com/, last checked May 30, 2007. Grosjean, D.; Grosjean, E.; Gertler, A. W., “On-road emissions of carbonyls from light-duty and heavy-duty vehicles” Environ. Sci. Technol., 2001, 35, 45-53. Harder J. W., Fried A., Sewell S., Henry B., “Comparison of tunable diode laser and long path ultraviolet/visible spectroscopic measurements of formaldehyde concentrations during the 1993 OH photochemistry experiment” J. Geophys. Res., 102, 6267-6282, 1997. Harrison R.M., Peak J.D., Collins G.M., “Tropospheric cycle of nitrous acid” Journal of Geophysical Research, 101, 14429–14439, 1994. Jang J.C., Jeffries H.E., Byun D., Pleim J.E., “Sensitivity of ozone to model grid resolution - I. detailed process analysis for ozone chemistry” Atmospheric Environment, 29, 3085-3100, 1995. Jang J.C., Jeffries H.E., Tonnesen S., “Sensitivity of ozone to model grid resolution - II. detailed process analysis for ozone chemistry” Atmospheric Environment, 29, 3101-3114, 1995. Jeffries H. E., Tonnesen S., “A Comparison of two Photochemical Reaction Mechanisms Using Mass Balance and Process Analysis” Atmospheric Environment, 28, 2991-3003, 1994. Jeffries H. E., Kim, B.U., Jobes, D., Pennington, E, “Role of Modeling Assumptions in the Houston Mid-Course Review” HARC Project H121.8HRB, Houston Advanced Research Center Report, http://files.harc.edu/Projects/AirQuality/Projects/H012.2004/ 30
8HRB/H12-8HRBFinalReportUNC.pdf, 2005. Kelly T. J., Fortune C. R., “Continuous monitoring of gaseous formaldehyde using an improved fluorescence approach” Inter. J. Environ. Anal. Chem., 54, 249-263, 1994. Kirchstetter, T. W., “Measurement of nitrous acid in motor vehicle exhaust” Environ. Sci. Technol., 30, 2843-2849, 1996. Kurtenbach R., Becker K. H., Gomes J.A.G, Kleffman J., Lorzer J. C., Spittler M., Wiesen P., Ackermann R., Geyer A., Platt U., “Investigations of Emissions and Heterogeneous Formation of HONO in a Road Traffic Tunnel” Atmospheric Environment, 35, 3385-3394, 2001. Lammel G., Perner D., Warneck P., “Nitrous acid at Mainz: observation and implication for its formation mechanism” In: Restelli, G., Angeletti, G. (Eds.), Physicochemical Behaviour of Atmospheric Pollutants, Kluwer Academic Publishers, Dordrecht, pp. 469-476, 1988. Lammel G., Cape J.N., “Nitrous acid and nitrite in the atmosphere” Chem. Soc. Rev., 361-369, 1996. Li, S. M., “Equilibrium of particle nitrite with gas phase HONO: Tropospheric measurements in the high Arctic during polar sunrise” Journal of Geophysical Research 99, 25469-25478, 1994. Li J., Dasgupta P. K., Luke W. T., “Measurement of Gaseous and Aqueous Trace Formaldehyde. Revisiting the Pentanedione Reaction and Field Applications” Anal. Chim. Acta, 531, 51-68, 2005. Martinez, M., Harder, H., Brune, W., Di Carlo, P., WIlliams, E., Hereid, D., Jobson, T., Kuster, W., Roberts, J., Trainer, M., Fehsenfeld, F.C., Hall, S., Shetter, R., Apel, E., Riemer, D., Geyer, A. 2002, “The behavior of the hydroxyl and hydroperoxyl radicals during TexAQS 2000”, Abstract A12D0174, AGU Fall meeting, San Francisco, CA, EOS Transactions. Merritt, P. M., “Diesel Exhaust Standard Phase 1: CRC Project NO. AVFL10A Final Report,” Prepared for Coordinating Research Council, Inc. by Southwest Research Institute San Antonio, TX, August 2003. MM5 Community Model 2007, http://www.mmm.ucar.edu/mm5/, last checked May 30, 2007. Neftel A., Blatter A., Hesterberg R., Staffelbach T., “Measurements of concentration gradients of HNO2 and HNO3 over a semi-natural ecosystem” Atmospheric Environment, 30, 3017-3025, 1996. Perner D., Platt U., “Direct measurements of atmospheric formaldehyde, nitrous acid, ozone, nitrogen dioxide, and sulfur dioxide by differential optical absorption in the near UV” J. Geophys. Res., 85, 7453-7458, 1980. Schmitz T., Hassel D., Weber F., “Determination of VOC-components in the exhaust of gasoline and diesel passenger cars” Atmospheric Environment, 34, 4639–4647, 2000. Seebold J. G.,Davis B. C., Gogolek P. E. G., Kostiuk L. W., Pohl J. H., Schwartz R. E., Soelberg N. R., Strosher M., Walsh P.M., “Reaction Efficiency of Industrial Flares: The Perspective of the Past” Petroleum Research Forum’s Project 92-19, 1997. 31
Seinfeld J., Pandis S. N., “Atmospheric Chemistry and Physics: From Air Pollution to Climate Change” Wiley-Interscience, 2006. Sigsby J. E., Tejada S., Ray W., “Volatile organic compound emissions from 46 in-use passenger cars” Environ. Sci. Technol., 21, 466-475, 1987. Siegl W. O., Hammerle R., Herrmann H., Wenclawiak B., Luers-Jongen B.,“Organic emissions profile for a light-duty diesel vehicle” Atmospheric Environment, 33, 797–805, 1999. Stutz J., Platt U., “Improving long-path differential optical absorption spectroscopy with a quartz-fiber mode mixer” Appl. Opt., 36, 1105-1115, 1997. Stutz J., Alicke B., Ackermann R., Geyer A., White A., Williams E., “Vertical profiles of NO3, N2O5, O3, and NOx in the nocturnal boundary layer: 1. Observations during the Texas Air Quality Study 2000” J. Geophys. Res., 109, Art. 12306, 2004. Texas Air Quality Study 2000, http://www.utexas.edu/research/ceer/ texaqs/ Texas Commission on Environmental Quality 2006, Houston-Galveston- Brazoria 8-Hour Ozone SIP Modeling, Available at http://www.tceq.state. tx.us/implementation/air/airmod/data/hgb2.html Texas Commission on Environmental Quality, http://www.tceq.state.tx. us/cgi-bin/compliance/monops/daily_summary?cams=603 Wert B. P., Fried A., Henry B., Cartier S., “Evaluation of inlets used for the airborne measurement of formaldehyde” J. Geophys. Res., 107, 4163, 10.1029/2001JD001072, 2002. Wert B. P., Trainer M., Fried A., Ryerson T. B., Henry B., Potter W., Angevine W. M., Atlas E., Donnelly S. G., Fehsenfeld F. C., Frost G. J., Goldan P. D., Hansel A., Holloway J. S., Hubler G., Kuster W. C., Nicks D. K., Jr., Neuman J. A., Parrish D. D., Schauffler S., Stutz J., Sueper D. T., Wiedinmyer C., Wisthaler A., “Signatures of terminal alkene oxidation in airborne formaldehyde measurements during TexAQS 2000” J. Geophys. Res., 108, Art. 4104, 2003. Wiedinmyer C., Guenther A., Estes M., Strange I., Yarwood G., Allen D. T., “A land use database and examples of biogenic isoprene emission estimates for the state of Texas, USA” Atmospheric Environment 35, 6465-6477, 2001. Zhu J., Cao X., Pigeon R., Mitchell K., “Comparison of vehicle exhaust emissions from modified diesel fuels” J. Air Waste Manag. Assoc., 53, 67-76, 2003.
32
