We report aerosol simulations using the EPA's Models-3 Community Multiscale Air Quality model (CMAQ) and ground-based and aircraft aerosol measurements to investigate new particle formation in Houston, Texas. The aerosol measurements reveal elevated ultrafine particles that reach the highest value in the afternoon, indicating prominent new particle formation. Simulations of the binary H2SO4-H2O nucleation predict an order of magnitude lower concentrations for aerosols near 10 nm than the measurements. A parameterized nucleation scheme that accounts for the enhanced nucleation effect of secondary condensable organics is incorporated into the Models-3/CMAQ. The organic nucleation scheme predicts the number concentrations in agreement with the measurements during the daytime. The diurnal variation is well reproduced in the simulations including the organic nucleation scheme. Comparison with the aircraft measurements also shows that the organic nucleation scheme produces good predictions of the altitude-dependent number size distributions of the ultrafine particles. The results corroborate the importance of secondary condensable organics in new particle formation when sulfate and organics are abundant.
 Aerosol particles have profound implications in the Earth-Atmosphere system. Aerosols directly and indirectly affect the Earth's energy balance [Houghton et al., 2001]. Aerosols also potentially impact air chemistry, human health, visibility, and weather [Orville et al., 2001; Zhang and Zhang, 2005]. Understanding these effects requires detailed information on how aerosol particles are formed and how they are transformed in the atmosphere. Key processes are the formation of new atmospheric particles and their subsequent growth to larger sizes.
 A variety of mechanisms have been proposed to explain aerosol nucleation in the atmosphere. The most widely studied one involves the binary water-sulfuric acid nucleation [Kulmala et al., 2004]. However, the predicted binary sulfuric acid-water nucleation rates are many orders of magnitude lower than observed in atmospheric nucleation events [Kulmala et al., 1998]. Ternary nucleation, involving the additional ubiquitous species ammonia, has been suggested from theoretical considerations as an explanation for observed particle nucleation in the atmosphere [Weber et al., 1999]. Other possible routes of atmospheric particle formation include ion-induced nucleation [Yu and Turco, 2001; Lee et al., 2003]. More recently, it has been suggested that condensable organic acids, produced from atmospheric oxidation of hydrocarbons, play an important role in particle nucleation [Zhang et al., 2004a].
 Sulfuric acid has been clearly identified as a major atmospheric nucleating species [Kulmala et al., 2004; McMurry et al., 2005]. Observed nucleation rates in the atmosphere, especially in urban areas, cannot always be explained by the nucleation of sulfuric acid with associated inorganic compounds (water and ammonia) [Yu, 2006]. Organic compounds having a low saturation vapor pressure (such as organic acids) appear to be likely candidates for particle nucleation [Zhang et al., 2004a]. Observation studies indicate the enhanced production of ultrafine particles in plumes rich in both H2SO4 and VOCs [Alam et al., 2003; Kulmala et al., 2004]. In this paper, we present simulations using the Models-3/CMAQ and ground-based and aircraft aerosol measurements to investigate new particle formation in Houston, Texas. A parameterized nucleation scheme that accounts for the enhanced nucleation effect of secondary condensable organics [Zhang et al., 2004a] is incorporated into the Models-3/CMAQ. Simulations with the organic nucleation scheme and binary H2SO4-H2O nucleation [Kulmala et al., 1998] are compared with the ground-level and aircraft measurements. Implications of the simulations on atmospheric new particle formation are discussed.
2. Model and Measurement Descriptions
 The Models-3/CMAQ system has been recently employed to simulate fine particulate matter (PM2.5) in Houston, Texas [Fan et al., 2005], and only a brief description of the modeling procedures relevant to this work are presented here. The Models-3/CMAQ Version 4.3 model system used in the present work was built on the Linux platform using Portland Group FORTRAN 90 (PGF90) compilers. The chemical mechanism chosen in this study was RADM2 (Regional Acid Deposition Model 2), with the aqueous chemistry extension. The model domain was comprised of 60 × 60 Lambert Conformal grids encompassing an area of 57,600 km2 with a 4-km resolution centered at 29.83 N and 95.05 W. There were 21 vertical layers from the surface to the top (50 mb), identical to those of gridded emission data. The selected simulation period spanned from August 22 to 29, 2004, corresponding to an 8-day episode. The meteorological fields for chemical transport simulations were generated using the Mesoscale meteorological Model version 5 (MM5) and post-processed by Models-3 meteorology-chemistry interface processor (MCIP2.2). The emission inventory used in this study was from EPA's National Emission Inventory (NEI99) final version 3. The raw NEI99 data were processed using Sparse Matrix Operator Kernel Emissions (SMOKE) Modeling System Version 1.5 β to obtain the gridded emission inventory data sets, ready for CMAQ runs [Fan et al., 2005]. Numerous measurement and modeling studies, particularly those associated 2000 Texas Air Quality Study (TexAQS) [Lei et al., 2004; Li et al., 2005; Zhang et al., 2004b], have documented the underestimation of the VOC emission inventory in the original NEI99, which needs to be modified for the modeling purpose in the Houston area. In the present work the emission inventory for aromatic hydrocarbons and olefins used in the simulations was eight times higher than that in the original NEI99, which was in accordance with the currently recommended emission inventory from the previous field and modeling studies.
 The classical binary H2SO4–H2O nucleation model parameterized by Kulmala et al.  was employed in the CMAQ aerosol module (referred to as the BN Scheme hereinafter). Alternatively, a nucleation scheme that accounted for the co-nucleation effects of H2SO4 and secondary condensable organics [Zhang et al., 2004a] was incorporated into the CMAQ aerosol module (refereed to as the ON scheme hereinafter). The nucleation rate (Jr) in the ON scheme was assumed to be dependent of the partial pressures of sulfuric acid and secondary condensable organics, with a parameterized equation, Jr = C · · (rorg,a + rorg,b), where rorg,a = Pi.,org and rorg,b = Pi.,org. Pi.,org denotes the concentrations of condensable organic species considered in the CMAQ aerosol module, and a and b represent the contributions from anthropogenic sources and biogenic sources, respectively. There were totally ten secondary condensable organics, consisting of primarily organic acids produced from oxidation of alkanes, olefins, aromatics, and terpenes. The concentrations of sulfuric acid and secondary condensable organics were predicted by CMAQ, with the typical concentrations of ∼107 cm−3 for sulfuric acid and ∼1012–1013 cm−3 for secondary condensable organics. C was a constant inferred from the above parameterized equation using the typical nucleation rate (∼106–107 m−3 s−1) measured in the urban atmosphere [Kulmala et al., 2004] and the CMAQ predicted concentrations of H2SO4 and secondary condensable organics, with a value of 3 × 10−25 m3 s−1. Considering the large uncertainties in our understanding of atmospheric new particle formation in experimental, theoretical, and modeling studies, such a parameterization represents a reasonable approach to account for the organic effect in new particle formation.
 Ground-level measurements of the aerosol number and size distributions were made during the simulation period using a differential mobility analyzer (DMA) system [Gasparini et al., 2004]. The ground site was located in Northeast Houston (Aldine, N29.90 and W95.33). The aircraft aerosol data were collected using a DMA on board the instrumented Southern Ogallal Aquifer Rainfall (SOAR) Cheyenne II aircraft during a flight on August 23, 2004.
Figure 1 shows a comparison of the predicted number size distributions from the two schemes with ground-level aerosol measurements at the Aldine Site, which were averaged over the daytimes (from 8 am to 6 pm) during the 8-day episode. The aerosol number size distributions averaged over the entire Houston area and over the daytimes during the episode from the two schemes were similar to those shown in Figure 1. There existed a significant difference in the number concentrations for particles smaller than 50 nm between the two schemes. The ON scheme predicted over one order of magnitude higher number concentrations than the BN for the nucleation mode particles (<11 nm). Clearly, the BN scheme was unable to simulate the measured high number concentrations of ultrafine particles, while the ON scheme yielded concentrations in good agreement with the measured values. For the nucleation mode particles (about 10 nm), the number concentration (dN/dlog(Dp)) predicted by the ON scheme was about 4.0 × 104 cm−3, which compared reasonably with the measured value of about 2.0 × 104 cm−3. On the other hand, the value predicted by the BN scheme was about a factor of twenty lower than the measured value.
Figure 2 presents the diurnal variations of measured and simulated number size distributions averaged over the eight simulation days. The measurements revealed high concentrations of ultrafine particles during the afternoon hours (Figure 2c). For the particles in the size range of 10–30 nm a maximum in dN/dlog(Dp) of greater than 3 × 104 cm−3 was reached between 1200 and 1700 local time, suggesting a prominent new particle formation event. The occurrence of this event is similar to that typically observed for ozone resulting from photochemical oxidation of volatile organic compounds in the presence of nitrogen oxides. The simulation employing the BN scheme predicted very low aerosol number concentrations throughout the day, especially for ultrafine particles (<30 nm) (Figure 2b). The predicted number concentrations of ultrafine particles with the ON scheme were considerably increased during the daytime (Figure 2a), which was in better agreement with the observations (Figure 2c). The results imply that co-nucleation between sulfuric acid and condensable secondary organics likely accounts for the observed new particle formation event during the afternoon hours. Because of very low H2SO4 and secondary organics concentrations during the night, the ON scheme somewhat underestimated nighttime ultrafine particle number concentrations. Nucleation in the nighttime was not controlled by H2SO4. It is likely that certain semi-volatile species (such as oxygenated VOCs) become more condensable as the temperature decreases at nighttime [Laaksonen et al., 2005].
Figure 3 shows the diurnal variations of predicted and observed total number concentrations at the Aldine site averaged over the episode. During the daytime, the particle number concentrations predicted by the ON scheme were in good agreement with the measurements, which were over 104 cm−3. The BN scheme considerably underestimated the particle number concentrations during the daytime, especially in the afternoon, by a factor of about 10. As discussed above, the ON scheme did not capture the observed nighttime nucleation and could not account for the nighttime particle concentrations.
 A comparison of the number size distributions predicted by the two schemes with aircraft measurements is presented in Figure 4. The aircraft data were averaged over a flight path to the north of Houston (from N30.135, W95.760 to N30.14, W 95.05) at the altitudes of 630–650 m. The model results for comparison were interpolated values accounting for the corresponding locations and altitudes. The predictions when using the ON scheme agreed reasonably with the aircraft measurements, but use of the BN scheme predicted the number concentrations of ultrafine particles substantially smaller than those measured.
 In this work we did not consider ternary nucleation of sulfuric acid-ammonia-water since a recent kinetic nucleation model constrained by experiments suggested a negligible role in new particle formation in the boundary layer atmosphere [Yu, 2006]. Furthermore, aerosol simulations and observations in the Houston area revealed dominant mass concentrations of sulfate and organics, but little nitrate in PM2.5 [Fan et al., 2005]. Also, we did not consider ion-induced nucleation. Recent measurements suggested that under fairly typical continental ground level conditions there was no evidence that either ion-induced or assisted nucleation were able to make a significant contribution to total particle production or growth [Eisele et al., 2006].
 Aerosol simulations using Models-3/CMAQ and ground-level and aircraft aerosol measurements have been presented to investigate new particle formation in Houston, Texas. The measurements indicated new particle formation events during the afternoon hours are frequent in the Houston area, correlating with photochemical VOC oxidation. The BN scheme was unable to simulate the measured high number concentrations of ultrafine particles. A parameterized nucleation scheme (ON scheme) that accounts for the co-nucleation effect of sulfuric acid and secondary condensable organics has been incorporated into the Models-3/CMAQ. The ON scheme predicted number concentrations of ultrafine particles in agreement with the measurements and reasonably reproduced the measured diurnal variation. Comparison with the aircraft measurements also showed that the ON scheme produced good predictions of the altitude-dependent number size distributions of ultrafine particles. The results support that secondary condensable organics are important in new particle formation when sulfate and organics are abundant.
 The authors are grateful to C.-J. Lin of Lamar University for providing the gridded emission data from NEI99 used in the simulations and Y. Lee and J. Santarpia of Texas A&M University for help with the data from ground level and aircraft aerosol measurements. This study was supported by NSF (ATM-0424885) and the Texas Air Research Center. J. Fan was supported by NASA Earth System Science (ESS) Fellowship.