A regional chemical transport model (CTM) has been employed to assess the influence of biogenic emissions on ozone (O3) formation in Houston, Texas during an eight-day episode of 24–31 August 2000 in association with the Texas Air Quality Study (TexAQS) 2000. The effects of isoprene and monoterpene emissions on O3 photochemistry in this region were investigated. Isoprene emissions played an important role in O3 formation when the O3 plume occurred in the afternoon over the urban Houston area. When the isoprene emissions were decreased or increased by 50%, the O3 concentration was decreased or increased by 5–25 ppb over the urban Houston area, respectively, but the change in the O3 concentration was less than 5–10 ppb over the industrial Ship Channel. Additional sensitivity studies showed that the surface O3 change resulted primarily from local isoprene emissions, although transport of isoprene from the north of the urban Houston area was found to be nonnegligible in the isoprene budget on several days. The contribution of monoterpene emissions to O3 formation was insignificant due to low emission rate and relatively slow reaction rate.
 Ozone (O3) is a key species in the tropospheric photochemistry and one of the criteria pollutants regulated by the US Environmental Protection Agency (US EPA). High O3 concentration in surface air is produced by rapid photochemical oxidation of volatile organic compounds (VOCs) in the presence of nitrogen oxides (NOx = NO + NO2). Elevated O3 concentrations are of major environmental concerns because of its adverse effects on human health [Lippman, 1993] and on ecosystems [National Research Council, 1991]. Air pollution continues to be a critical problem in major urban centers around the globe. Air pollution in the Houston area is noted for some of the highest O3 concentrations encountered in the continental US. The 1-hour averaged daily ozone maximum exceeded 250 parts per billion (ppb), the highest value recorded in the USA over the past decade.
 Houston is the fourth largest metropolitan area in the United States. The presence of transportation facilities leads to high emissions of NOx and VOCs from mobile sources. Large amounts of vegetation and forest in the northeast of Houston may be the source of emissions of biogenic VOCs such as isoprene and monoterpene, which, when transported into the urban area and mixed with NOx from anthropogenic sources under favorable meteorological conditions, can contribute to the O3 formation in the urban area [Wiedinmyer et al., 2001; Kleinman et al., 2002]. Furthermore, in contrast to other large cities in the United States, Houston hosts one of the largest concentrations of petrochemical industries in the world, representing a rich source of NOx and highly reactive VOCs. These industrial facilities spread throughout the southeast Texas, and the highest density is located in the region surrounding the Houston Ship Channel.
 The contribution of biogenic emissions such as isoprene and monoterpene to the O3 formation in the Houston area is yet to be quantified. The role of biogenic emissions on surface O3 has been the subject of considerable debate since 1970s. Early studies suggested that the biogenic emissions in urban areas contribute negligibly to O3 formation because the biogenic emissions were relatively small compared with the anthropogenic emissions, on the basis of a review of sources, emission rates, ambient measurements of biogenic VOCs [Altshuller, 1983]. Trainer et al.  and Chameides et al.  pointed out the potential importance of biogenic VOCs to the formation of ground level O3 in a study of the reasons behind the limited successes of O3 abatement strategies. More recently, a number of studies focused on the role of naturally emitted VOCs on concentrations of surface O3 in North America [Sillman et al., 1990; McKeen et al., 1991; Pierce et al., 1998; Diem, 2000; Mendoza-Dominguez et al., 2000; Tao et al., 2003], in Europe [Simpson, 1995; Vogel et al., 1995; Müzenberg-St. Denis and Renner, 1999; Varinou et al., 1999; Andronopoulos et al., 2000; Thunis and Cuvelier, 2000; Toll and Baldasano, 2000; Vlachogiannis et al., 2000], and in Asia [Shao et al., 2000; Han et al., 2005]. The general conclusions drawn from these studies were that biogenic emissions enhanced O3 formation in most areas and naturally emitted VOCs, especially isoprene, played a significant role in ground-level O3 due to their relatively high reactivity [Fuentes et al., 2000]. Recent studies also provided the qualification of the contribution of biogenic emissions to the O3 formation in some urban areas. Solmon et al.  used the modeled biogenic emissions to study the impacts on regional O3 formation and suggested that the introduction of biogenic fluxes led to an increase in simulated surface O3 concentrations, reaching 18–30% in the Paris urban plume. In the study of Bell and Ellis , an increase by 100% in biogenic VOC emissions raised the O3 level, with an estimated maximum 1-hour concentration of 30% higher than that of the baseline scenario and with a larger increase in urban areas than rural regions in mid-Atlantic region. In the simulation of Cortinovis et al. , for the urban (Marseille) scenario, the impact of biogenic emission on O3 production was large and the enhancement of O3 production due to isoprene reached 37% in terms of maximum surface concentrations and 11% in terms of the total O3 production. In addition, biogenic emissions contribute to formation of secondary organic aerosols, which has important implications to urban air pollution and climate [Zhang et al., 2004a; Fan et al., 2006].
 The biogenic emissions in the Houston area are typically much less than those from anthropogenic and industrial sources, but there are large amounts of biogenic emissions in the northeast of Houston which can be transported to the urban area under favorable meteorological conditions to contribute to O3 formation. In this paper, a chemical transport model (CTM) was employed to assess the contribution of biogenic emissions to O3 formation in the Houston area. An eight-day episode during 24–31 August 2000, during which emission inventory and meteorological data were available, was considered in the present work. Budget analysis of isoprene was performed to evaluate the contributions of transport processes to isoprene and O3 concentrations. Furthermore, there are large uncertainties in estimates of biogenic emissions from inventories and emission algorithms [Environmental Protection Agency, 1996; Geron et al., 1995; Guenther et al., 2000; Kesselmeier and Staudt, 1999; Pierce et al., 1998; Potter et al., 2001; Roselle, 1994, Roselle et al., 1991; Wiedinmyer et al., 2001]. Considering the uncertainty in the estimate of isoprene emission inventory [Wiedinmyer et al., 2001; Song et al., 2004], sensitivity simulations were designated with increasing and decreasing isoprene emissions by 50% to determine the contribution of isoprene emissions to O3 concentrations.
 The CTM employed in the present work was initially developed at the National Center for Atmospheric Research (NCAR) [Hess et al., 2000] and modified for photochemical modeling in the Houston area [Lei, 2003; Lei et al., 2004; Zhang et al., 2004b]. Briefly, there were 38 layers in the vertical direction from the surface to 100 HPa in a terrain-following coordinate system with seven layers in the lowest 500-m altitudes. The chemistry in the CTM included treatment of standard gas-phase and heterogeneous chemistry [Lei et al., 2004]. The organic part was based on the CB-4 mechanism used in the Comprehensive Air Quality Model with the extension version 3.1 (CAMx v3.1) [Gery et al., 1989; Simonaitis et al., 1997; CAMx v3.1, 2000]. The CB-4 mechanism was modified to represent both polluted and remote tropospheric atmosphere [Lei et al., 2004]. Recently, the isoprene atmospheric chemistry has been substantially improved by experimental and theoretical studies [Fan and Zhang, 2004]. The kinetics and product yields of the OH-isoprene reactions have been investigated through experimental studies. Theoretical calculations employing quantum chemical methods and kinetic rate theories have predicted the reaction rate constants and isomeric branch ratios of the OH-isoprene reaction and several intermediate steps. These recent advances on isoprene oxidation mechanism were included in the CB-4 mechanism [Lei et al., 2000; Lei et al., 2001; Zhang and Zhang, 2002]. The heterogeneous reaction of N2O5 on sulfate aerosols was included in the model [Zhang et al., 2004b]. The chemical initial and lateral boundary conditions were interpolated from modeling ozone and related chemical tracers (MOZART, v2), a global CTM [Horowitz et al., 2003; Tie et al., 2003; Zhang et al., 2003] for the coarse grid run. The CTM was driven by the output circulation field from a meteorological model, the Penn State/NCAR Mesoscale Model, version 5 (MM5) [Grell, 1993]. The meteorological fields used in this study were from the MM5 output nudged by observational data [Nielsen-Gammon, 2002].
 An eight-day episode of 24–31 August 2000 was chosen for this study. This episode is in association with the TexAQS 2000. A primary objective of the TexAQS 2000 was to examine the effects of different sources of O3 precursors on the occurrence of O3 episodes in and around the Houston area. This episode corresponded to drier conditions than normal, leading to unusually high temperatures during the latter half of the episode. The meteorological conditions during the earlier half of the episode, 24 to 29 August, were characterized by large-scale onshore flow and stagnant or near-stagnant conditions during midmorning; in the afternoon, the wind increased from the southeast and rotated clockwise to develop a westerly component after midnight. The latter part of the episode, beginning on 30 August, was characterized by shore-parallel or offshore large-scale flow and stagnant or near-stagnant conditions during midafternoon. Pollutants were frequently transported southeast away from the city to the Galveston Bay, only to return behind the bay breeze later in the afternoon [Nielsen-Gammon, 2002].
 Model simulations were initially performed with a 12-km horizontal resolution and 90 × 90 grid points in the eastern Texas at 0000 UTC 24 August 2000, allowing 24 hours for spin-up to remove the influences of initial conditions. The chemical lateral boundary conditions, the upper boundary conditions, and the initial conditions were linearly interpolated in space and time from daily averaged concentrations of the output of MOZART. We subsequently conducted model simulations using a 4-km horizontal resolution with 84 × 65-grid points in the Houston area. The outputs with the 12-km resolution were used for the initial and boundary conditions in the simulations of the 4-km resolution. The meteorological field and emission inventory (EI) were provided by the Texas Commission of Environmental Quality (TCEQ) via a ftp site (ftp.tceq.state.tx.us). The surface EI was chemically speciated (based on the CB-4 mechanism), spatially gridded (4 × 4 or 12 × 12 km), and temporally resolved (1-hour averaged). Numerous measurement [Kleinman et al., 2002; Berkowitz et al., 2004; Daum et al., 2003; Ryerson et al., 2003; Daum et al., 2004; Karl et al., 2003; Wert et al., 2003] and modeling [Lei et al., 2004; Fan et al., 2005; Li et al., 2005; Zhang et al., 2004b] studies, particularly those associated with TexAQS 2000, have documented the underestimation of VOCs in the original TCEQ emission inventory (EI), which needs to be modified for the modeling purpose in the Houston area. In the present work the emission inventory for light olefins used in the simulations was three times higher than that in the original TCEQ EI, in accordance with the currently recommended emission inventory from the previous field and modeling studies [Ryerson et al., 2003; Karl et al., 2003; Wert et al., 2003; Jiang and Jerome, 2004]. The biogenic emissions from TCEQ were estimated using a new biogenic model, Globeis2 [Guenther et al., 1997]. In estimating biogenic emissions, species composition, leaf biomass density, land use/land cover (LULC) map and meteorological variable were taken into account [Wiedinmyer et al., 2001].
where Kc (m2 s−1) is the coefficient of vertical diffusivity and γc is the no nonlocal transport term, which represents transport due to large-scale boundary layer eddies that occur during convective and unstable conditions. γc and Kc were inputted from the MM5. Above the PBL γc was set to zero. The isoprene emission was treated as a lower boundary condition on the surface flux equation. Surface deposition of O3 was parameterized following Wesely . The PBL height used was from the output of MM5.
 We divided the Houston Metropolitan area into an urban region encompassing the city of Houston (the urban Houston) and an industrialized Ship Channel region to east of the city (the Ship Channel), to reflect distinct influences of biogenic emissions (Figure 1). The combined urban Houston and Ship Channel regions are referred to as the Houston area in this work. The distribution of isoprene emission averaged during the entire episode is depicted in Figure 1. The isoprene emission in the urban Houston and Ship Channel regions is much lower than VOC emissions from the industrial sources. However, there are large emissions of isoprene to the north of Houston, which may affect O3 formation in Houston when the meridional wind is from north to south.
 Several model sensitivity studies were performed, including (1) a base run with all anthropogenic and biogenic emissions; (2) a control run without isoprene emission; (3) a control run without monoterpene emission; (4) a control run without the isoprene emission outside of the Houston area; (5) a control run with the isoprene emission decreased by 50%; and (6) a control run with the isoprene emission increased by 50%. Budget analysis of isoprene was performed on the basis of the base run.
3. Results and Discussion
3.1. Ozone Simulations
Figure 2 depicts the spatial distributions of calculated and observed near-surface concentrations of O3 at 15 Central Daylight Time (CDT) from 25 to 31 August 2000. The spatial patterns of predicted and observed O3 were well correlated. On 25 August, stagnant conditions during the late morning allowed an area of high O3 to develop in the Ship Channel area and migrate westward during the day. At 15 CDT, the O3 plume was located to the northeast of the urban Houston according to the measurements; the predicted O3 plume moved to the middle of the urban Houston and CTM predicted the O3 concentrations consistent with the observation sites. On 26 August, midday winds were light from the southeast. The O3 plume moved outside of the urban Houston at 15 CDT and the predicted concentrations of O3 were in agreement with the measurements at most of the observation sites. On 27 and 28 August, noontime winds were well organized from the southeast and the O3 plume was formed downwind of the urban Houston and Ship Channel at 15 CDT. The simulated O3 concentrations were consistent with the measurement at all observation sites. On 29 August at the noontime, winds were generally light and disorganized throughout the Houston area. By the afternoon, winds were systematically from the southeast. The CTM well reproduced the O3 plume to the north of the urban Houston and over the entire Ship Channel region. On 30 and 31 August, winds during the morning were from the west or northwest and tended to be lightest during the afternoon. The highest O3 was measured during the late afternoon. The modeled distribution of O3 generally well correlated the measurements; however, O3 was somewhat underestimated at a few observation sites, especially in the urban Houston.
 The model performance is further illustrated in Figure 3a, showing the diurnal variation of modeled and observed near-surface hourly O3 averaged over all 18 monitoring sites in the Houston area. The model generally well tracked the O3 diurnal variability. For example, the rapid increase of morning O3 concentrations was well replicated, and the falloff of afternoon O3 levels and the low nighttime O3 concentrations were well simulated. However, the CTM underestimated the O3 concentration in the early afternoons on 30 and 31 August, which could be explained by uncertainties in the meteorological inputs. It has been suggested that the modeled bay breeze was delayed and lighter in MM5 simulations on those days [Seaman, 2000; Solomon et al., 2000; Nielsen-Gammon, 2002].
Figures 3b and 3c show a scatterplot of modeled and measured O3 concentrations for all monitoring stations during the daytime and nighttime of the weeklong simulation, respectively. Comparison between the simulated and measured O3 concentrations (ppb) yielded a best fit line of [O3]mod = 0.82[O3]obs + 13.8 with a correlation of R2 = 0.75 for the daytime and a best fit line of [O3]mod = 0.31[O3]obs + 8.9 with a correlation of R2 = 0.09 for the nighttime. The modeled O3 concentrations were generally in agreement with the measured values at daytime, although there existed deviations between the simulated and observed O3 on 25, 26, and 29–31 August at the Ship Channel sites. In general, the bay and sea breezes predicted by MM5 were slower and lighter than the observations during those periods [Nielsen-Gammon, 2002]. In Figure 3c, the simulation for nighttime O3 deviated from the observation because of difficulties in modeling the meteorological fields during the night and the complexity of the nighttime chemistry. With high emissions in the urban and industrial area, NOx rapidly accumulated and consumed O3 at the surface during the nighttime. For the simulated O3 concentration of more than 60 ppb, the mean bias and normalized mean bias for surface O3 predictions were 5.1 ppb and 15.2%, respectively.
 Measurements from the research aircraft provided detailed spatial and temporal information on the variation of O3 that could not be obtained form the surface sites. The simulated O3 is compared with measurements obtained from the G-1 research aircraft in Figure 4. CTM captured the O3 plume observed by the aircraft on 25 August, but the CTM overestimated the O3 concentration by about 30 ppb in the plume. On 26, 29, and 31 August, CTM well tracked the O3 variation except for somewhat underestimates of O3 concentrations in some plumes. On 30 August, the CTM reproduced the O3 plumes, but there were offsets between the model and measured O3 plumes at 2030 and 2130 CDT. The discrepancies between simulated and observed O3 during the two time periods were significant, and were mainly due to the uncertainties in model simulations related to the bay breeze, which might cause a postpone in the formation of O3 plume on 30 August. The mean bias and normalized mean bias for aloft O3 predictions were −1.5 ppb and 5.6%, respectively.
3.2. NOy and PAN Simulations
Berkowitz et al.  reported observations of O3, NO, NOy, CO, SO2, PAN, HNO3, and HCHO made on the 65th floor of the Williams Tower in western Houston (−95.474°W,+29.750°N) between 15 August and 15 September 2000. This station was located at an altitude of 250 m above the ground level and was deployed as part of the TexAQS 2000. Measurements from the Williams tower were recorded as 15-min average values. The simulated NOy and PAN are compared with the measurements from the Williams tower during the entire episode (Figure 5). The simulation for NOy in Figure 5a agreed well with the observation. For instance, the early morning peak NOy concentrations were well reproduced, except with some underestimates on 25, 26, and 29 August. The daytime NOy concentrations were also well simulated. CTM well tracked the variations of PAN at the Williams Tower, but underestimated at daytime on 29, 30, and 31 August. One possible reason for the underestimates in PAN was associated with the underestimates in O3 on those days. As to be discussed, when isoprene emissions were increased by 50%, the O3 simulation was improved on these days, and the PAN simulation was improved similarly (not shown). Alternatively, the PAN simulation could be related to the rapid vertical exchange resulted from the MRF PBL scheme.
3.3. Isoprene Simulations
Karl et al.  reported measurements of VOCs at La Porte (−95.050°W, 29.667°N) near the Houston Ship Channel during TexAQS 2000 by using proton-transfer reaction mass spectrometry (PTR-MS). The uncertainty of the VOC measurements was estimated on the order of 30% and the detection limits were around 20 pptv. The PTR-MS was not operated on 26 and 27 August, and the isoprene data was missing on these 2 days. When the isoprene measurement was available, the sample rate was about every 5 min. Compared with the limited isoprene measurements during TexAQS 2000, the first peak in the morning was captured by the CTM, but there existed deviations between the simulated and observed isoprene concentrations of the first peak on some days (Figure 6). For example, the first isoprene peak on 25 August was 1.8 ppb from the observation, but only 0.7 ppb from the CTM, with a difference of 61%. In addition, the model underestimated the second peak in the evening. As to be discussed, an increase in isoprene emissions by 50% (purple line in Figure 6) did not reduce the discrepancies, indicating that factors other than emissions might play a role. Because isoprene is a primary precursor of O3 with highly localized distribution from biogenic sources and has a very short lifetime in the PBL, the PBL height is critical to simulate the isoprene concentration in the CTM. The La Porte super site is located near the Galveston Bay, which is prone to be affected by the air over sea. The simulation of isoprene in the CTM was dependent on the PBL height, which was controlled by the air over land or sea. Nevertheless, the simulated isoprene is encouraging, considering the large uncertainties in the isoprene emissions and meteorological conditions.
3.4. Effect of Isoprene Emission on Ozone Formation
Figure 7 shows the change of O3 near the surface at 15 CDT due to the exclusion of isoprene emission in the CTM. The response of the ground level O3 to exclusion of isoprene emissions was remarkable from 29 to 31 August in the Houston area, especially in the urban Houston where O3 decrease was up to 40 ppb at 15 CDT. The influence of isoprene emissions on the O3 concentration was relatively minor on 27 August in the Houston area and the O3 reduction was less than 15 ppb.
 From the distribution of the O3 change depicted in Figure 7, the impacts of isoprene emissions on the O3 plume from different source regions in the Houston area are distinct. The Parrish Power Plant, in the southeast of Houston, emits large amounts of NOx but little VOCs, making the VOC/NO2 reactivity ratio in the plume very low and requiring the admixture of ambient air containing VOCs to efficiently produce O3. As a result, when isoprene emission was removed, the O3 concentration in the downwind of the plant was decreased noticeably, with the maximal O3 reduction of more than 40 ppb. Similar to most urban areas in the United States, emissions from the urban Houston originate primarily from mobile sources and the VOC/NO2 reactivity ratio is fairly low. The O3 production rate and efficiency are VOC limited. The effect of isoprene emission on the O3 plume from the urban Houston is apparent, but less than that from the Parrish Power Plant. Emissions from the Ship Channel region have VOC/NO2 reactivity ratios that are many times higher than those from the urban Houston, representing the conditions which promote rapid and efficient formation of high concentrations of O3. The decrease of O3 due to the exclusion of isoprene emissions was rather insignificant.
Figures 8a and 8b show the impacts of isoprene emissions on O3 diurnal variation averaged over the observational sites in the urban Houston and Ship Channel, respectively. From 25 to 28 August, the changes in the O3 diurnal cycle were relatively minor in both regions. However, from 29 to 31 August the isoprene emission played a significant role in O3 formation in the urban Houston and the peak O3 concentration decreased by about 10 to 20 ppb. In contract, the O3 decrease was less than 10 ppb during the same period in the Ship Channel. The changes in the O3 diurnal cycle in these two regions were consistent with the corresponding VOC/NO2 reactivity ratios.
3.5. Isoprene Budget and Sensitivity Studies
 To assess the impact of isoprene emissions on O3 formation in the Houston area, we analyzed the budget of isoprene, emphasizing on the various processes affecting the isoprene concentration. We averaged the isoprene concentration over the Houston area as shown in Figure 1 (the urban Houston and Ship Channel) and within the PBL. The net changes in the isoprene concentrations due to horizontal transport, vertical transport (including advection, diffusion, and PBL mixing), photochemistry, emissions, and dry deposition were summed over the grid points and over 8 hours from 10 to 18 CDT to derive an average change for each individual process during the entire episode. The analysis showed that the amount of isoprene within the PBL was primarily controlled by the surface emission and photochemical loss except on the last 2 days, during which north-south advection and vertical transport had significant contributions to the isoprene concentration. As discussed in section 3.3, the maximal O3 reduction due to elimination of isoprene emissions occurred on the last 2 days. Hence transport of isoprene from the outside area of the urban Houston was of importance in the O3 formation and the wind direction was an important factor, as previously suggested by Kleinman et al. . As demonstrated in Figure 1, a large amount of isoprene is emitted to the north of the Houston area. When the meridional wind was from north to south, transport of isoprene into the urban Houston area was important to the isoprene concentration.
 A sensitive study was conducted to investigate the contributions of isoprene transport from outside of the Houston area to O3 formation. In this case, the isoprene emission outside of the Houston area was removed. The diurnal variations of O3 concentrations over all monitoring sites in these two cases were similar, except for the last 2 days when the difference was about a few parts per billion (not shown). Therefore the contributions of isoprene transported from outside of the Houston area were insignificant to O3 formation at the surface, which may be explained because of the short lifetime of isoprene within the PBL. However, as discussed above, north-to-south transport of isoprene can be important to the isoprene budget in the Houston area.
 There are large uncertainties in estimates of biogenic emissions from inventories and emission algorithms [Wiedinmyer et al., 2001; Song et al., 2004], although Figure 6 does not appear to suggest that the isoprene emission inventory might be underestimated except on 25 and 26 August at La Porte (in the Ship Channel, Figure 1). Berkowitz et al.  found the reactivity at the Williams Tower (in the urban Houston, Figure 1) at 1300 LST was dominated by isoprene from their measurements between 15 August and 15 September 2000 and the mean isoprene reactivity was 1.44. In the model simulation and observations, the reactivity of isoprene was defined as KOH [isop], where KOH is the reaction rate constant between OH and isoprene and [isop] represents the isoprene concentration. The simulated mean isoprene reactivity at 1300 LST at the Williams Tower during the whole episode was 0.92, lower than the measurement. In order to evaluate the contribution of isoprene emissions to O3 concentration, two additional sensitivity simulations with the isoprene emission increased and decreased by 50% were performed. When the isoprene emissions were decreased by 50%, the response of the ground level O3 was significant between 29 and 31 August in the urban Houston, where the O3 concentration was decreased by 5–25 ppb at 15 CDT. However, the decrease of O3 concentration in the Ship Channel was small, less than 10 ppb. When the isoprene emissions were increased by 50%, the increase in O3 concentration between 29 and 31 August in the urban Houston was about 5–20 ppb at 15 CDT. In the Ship Channel, however, the increase of O3 concentration was negligible between 29 and 31 August, less than 5 ppb. Figure 8 shows a comparison of the diurnal variation of O3 concentration with adjusted (±50%) isoprene emissions in the Houston area. The simulated O3 varied little due to adjusted isoprene emissions from 25 to 29 August in the urban Houston and Ship Channel. However, during 30 and 31 August with the increased and decreased (±50%) isoprene emissions the peak O3 concentration increased and decreased by about 10 ppb in the urban Houston and 5 ppb in the Ship Channel, respectively. The O3 simulation with the isoprene emissions increased by 50% was improved compared with the measurements and the simulated mean isoprene reactivity at 1300 LST at the Williams Tower was 1.30, close to the measurement on 30 and 31 August.
 Sensitivity studies were performed to evaluate the role of monoterpenes in O3 formation in the Houston area. In all cases, the contribution of monoterpene emissions to O3 formation was negligible, likely because of their low emission rates and slow reaction rates. The O3 reduction due to exclusion of monoterpene emissions was less than 2 ppb.
 In this study a three-dimensional regional chemical transport model has been applied to assess the impacts of biogenic emissions on O3 formation in the Houston area during TexAQS 2000. The model performance was evaluated by comparing simulated O3 concentrations to those observed by a surface-monitoring network and the G-1 research aircraft. The magnitude, location, and movement of O3 plumes were well reproduced and the temporal and geographical variations of O3 were well simulated. The predicted and measured concentrations of NOy at the Williams Tower were in agreement. CTM well tracked the variations of PAN at the Williams Tower, but the simulated concentrations of PAN on 29–31 August were lower than the observations. The simulated isoprene compared reasonably with the available observation, considering the uncertainties associated with isoprene emissions and meteorological inputs.
 Isoprene emission played an important role in the O3 formation in the Houston area when the O3 plume occurred in the urban Houston in the afternoon. The reduction in the O3 concentration was up to 20–40 ppb due to exclusion of isoprene emissions in the urban Houston from 29 to 31 August. The VOC/NO2 reactivity ratio determined the influence of isoprene emissions on O3 formation. In the downwind of a major power plant and in the urban Houston, the VOC/NO2 reactivity ratio was low and the contribution of isoprene emissions was significant. In contrast, the VOC/NO2 reactivity ratio was much higher in the Ship Channel and the contribution of isoprene emission was relatively minor.
 Although there was a large amount of isoprene emissions to the north of Houston and the budget analysis also demonstrated transported isoprene into the urban Houston on certain days, the sensitive study showed that the surface O3 concentration in the Houston area was reduced by a few parts per billion when the isoprene emission outside of Houston was removed. The local emission of isoprene played a key role in the O3 change in the Houston area. O3 simulation was improved and the isoprene emissions were increased by 50%. It should be pointed out that there are likely other factors which also affect the performance of CTM, such as the grid resolution, chemical mechanism, meteorological input, and emission inventory.
 Finally, the contribution of monoterpene emissions to O3 formation in the Houston area was unimportant because of low emission rates and slow reaction rates.
 The authors would like to thank Thomas Karl for providing the isoprene measurement at La Porte and Carl M. Berkowitz for providing the NOy and PAN measurements at Williams Tower. This study was partially supported by NSF (ATM-0424885) and the US Environmental Protection Agency EPA (R03-0132). J. Fan was supported by a NASA ESS fellowship. The National Center for Atmospheric Research (NCAR) is supported by the National Science Foundation (NSF).