3.1. OH Measurements and Model Comparison
 The results of OH, H2SO4, and MSA measurements performed aboard the P-3B during TRACE-P can be found in the GTE data archive at http://www-gte.larc.nasa.gov/trace/TP_dat.htm. Data are presented as 30 s measurements using the same data acquisition technique as described by Mauldin et al. . As in previous airborne studies, most of the large scale variability in the data is due to changes in altitude. Smaller scale features which are observed during constant altitude legs are due to changes in [H2O], j(O3 O(1D)), and/or changes in [NO].
 Photochemical box model simulations of OH data were performed using a model developed at the Georgia Institute of Technology and enhanced at NASA Langley. A full discussion of the model used for these OH simulations and its use for the effects of clouds upon other radical species can be found in the work of [Crawford et al., 2003]. The reader is directed there for model details. A typical comparison of measured and modeled [OH] for a single flight can be found in Figure 3. Also shown is a plot of [NO], a key species in controlling the cycling of HO2/RO2 into OH. A relative altitude profile has been added for reference. As can be seen, there are regions where the model agrees well with measurements and others where it does not. Most of the overall fine structure observed in the measurements is also seen in the modeled values. An interesting feature on this plot is the large spike of NO, from a ship plume encounter, which occurs at ∼0210 UTC. When NO concentrations increase to levels up to ∼150 parts per trillion by volume (pptv), OH concentrations are enhanced due to increased cycling of HO2 back into OH. However, at NO concentrations above ∼150 pptv, the accompanying NO2 begins to act as a sink for OH via pathways that lead to the formation of HNO3 (and HO2NO2). From the inset in Figure 3 it can be seen that during the NO spike the measured OH increases by ∼50%, going from ∼4 × 106 to ∼6 × 106 molecule cm−3. This modest increase is a combination of the enhanced cycling of HO2 into OH and a larger sink for OH via reactions with NO2 (and additional hydrocarbons). The flight path of the aircraft tried to follow the plume, encountering it from the up wind direction. It can be seen that following the initial encounter OH drops ∼6 × 106 to ∼2 × 106 molecule cm−3 and then begins a slow rise back to the value of ∼4 × 106 molecule cm−3 observed before the plume encounter. This drop is due to increased losses of OH via reactions with hydrocarbons and NO2, which then begin to dilute, allowing OH to climb back to its preplume value. While not agreeing exactly with the measurements, the model also predicts a similar increases and decrease in OH in this highly chemically perturbed environment. A plot of measured versus modeled [OH] for flights 4–24 (all flights during the TRACE-P campaign) is shown in Figure 4. As can be seen with the exception of a few outlying points, the agreement is generally good. A linear regression with a zero intercept of this data yields a slope of 0.89. A standard regression yields a slope of 0.85 and an intercept of 5.22 × 105 molecule cm−3. The slope of either regression indicates that the model has a slight tendency to overpredict OH concentrations when compared to the present measurements. This type of figure is good at revealing average overall trends; however, it is not good at elucidating specific types of dependencies. A cursory look at the agreement between measured and modeled values in Figure 3 reveals that the modeled values are lower than measurements at low altitudes and larger at high altitudes. Figure 5 is a plot of the ratio of measured to modeled [OH] versus altitude. As can be seen, at sea level the ratio is on average ∼1.5, indicating that the model is underestimating [OH] compared to measurements. As altitude increases the average ratio drops to unity at ∼0.8 km and then falls to a more or less constant value of ∼0.8 at altitudes >2 km. It is this large number of measurements made above 2 km that is dominating the fewer measurements below 2 km to give an overall agreement of 0.85–0.89, depending upon the regression used.
Figure 3. Plot of measured and modeled OH concentrations for a typical flight together with observed [NO], a key specie in controlling OH. A relative altitude profile (0.2–5.3 km) is provided for reference. As can be seen much of the high-frequency changes are due to changes in [NO]. The large spikes in NO (due to a ship plume) are large enough to suppress OH by acting as a sink to HOx (see text).
Download figure to PowerPoint
Figure 4. Comparison of measured OH values from the entire TRACE-P study together with those obtained from model simulations. A linear fit to this data yields a slope of 0.85 with a correlation coefficient of 0.84, indicating an overall tendency for the model to overestimate measurements.
Download figure to PowerPoint
Figure 5. Plot of the ratio of measured to modeled OH versus altitude. As can be seen, the model tends to underpredict measured OH at altitudes <1.5 km. At altitudes >1.5 km the model tends to overestimate OH measurements. As the bulk of the measurements were made at altitudes >1.5 km, this overestimation dominates the overall model agreement seen in Figure 4.
Download figure to PowerPoint
 Putting the present overall trend value of 0.85 (the slope from the same type of regression as used in these studies) into historical perspective, it compares well with the central pacific values of 0.86 and 0.80 obtained during the PEM-Tropics B study by Tan et al. [2001a] and Mauldin et al. , respectively. This contrasts to the generally poor agreement with the continental values of 0.7, 0.5, and 0.38 obtained by Carslaw et al. , Carslaw et al. , and Tan et al. [2001b], respectively. When compared to the values obtained in the TOPSE study, the present value agrees with the value of 0.8 obtained at higher latitudes (>70°N) but does not agree very well with the value of 0.6, obtained at lower latitudes (<57°N) [Mauldin et al., 2003]. The altitude trend of underestimating OH at low altitudes and overestimating OH at high altitudes in the model agreement has also been observed in previous studies [Mauldin et al., 1999, 2001; Tan et al., 2001a] and is also present in model simulations of OH data obtained aboard the DC-8 during TRACE-P [Crawford et al., 2003]. An intercomparison of OH and other data obtained from both the P-3B and DC-8 aircraft during three close proximity flight legs can be found in the work of Eisele et al. .
 As pointed out by Mauldin et al. , Tan et al. [2001a, 2001b], and Carslaw et al. [2001, 1999, and references therein], the tendency for models to overestimate OH concentrations seems to be enhanced for measurements performed in continentally influenced air. One historically common explanation is the presence of one or more unaccounted for hydrocarbon compounds, perhaps oxygenated species, which act as OH sinks [McKeen et al., 1997]. In continentally influenced regions, where biogenic carbon production is larger and transport of hydrocarbons and other compounds from local sources can increase the hydrocarbon loading, models begin to overestimate OH concentrations. What is interesting in this study is that while the measurements were performed over water, they were obtained in regions of continental outflow. One possibility for the better agreement in this region of continental influence, compared to other studies, is the addition of HO2 and HO2 + RO2 measurements [Cantrell et al., 2003] to the model. These two radicals are key species in the HOx cycle, and the addition of their measurement is expected to improve OH simulations.
 One other possibility is that the lifetime(s) of this/these hypothesized unaccounted for hydrocarbon(s) is sufficiently short that they are substantially reduced or removed during transport from the continent to out over the water. A quick test of the feasibility of this possibility can be performed by assuming pseudo-first-order conditions in which the [OH] remains constant and then estimating an equivalent rate coefficient for reaction with OH necessary to remove a hydrocarbon representative of the unaccounted for hydrocarbon specie(s) during transport. Here the lifetime of the representative hydrocarbon is the reciprocal of the apparent rate coefficient (the product of the [OH] and the second-order rate coefficient). A typically observed midday OH concentration of 3 × 106 molecule cm−3 equates to ∼1 × 106 molecule cm−3 when the diurnal cycle is averaged over 24 hours. If one assumes a transport time (hydrocarbon lifetime) from the continent to the measurement region of 0.5–1.5 days (reasonable for this study), second-order rate coefficients of 2.3 × 10−11–7.7 × 10−12 cm3 molecule−1 s−1 are obtained for reaction with OH. These values fall in the range of rate coefficients for OH reacting with formaldehyde and OH reacting with propane, ∼1 × 10−11 and ∼2.4 × 10−13 cm3 molecule−1 s−1, respectively. Thus it is possible that once the air mass has moved from the continental source regions, OH could have substantially reduced unmeasured hydrocarbon species during transport and thus their impact upon steady state OH concentrations in the measurement regions.
3.3. Nighttime Flight Off Midway Island
 An example of the precision the instrument is capable of when measuring low OH is P-3B flight 21, flown in the dark out of Midway Island. These types of flights are also good checks for possible interferences as OH concentrations are typically <105 molecule cm−3 during the night for remote environments. Figure 7a is a plot of the observed [OH] during this flight. Shown are the individual 30 s measurements and 5 min averages. Also shown is an altitude profile for reference. With the exception of some noisy excursions, it can be seen the 30 s values are <2 × 105 molecule cm−3, a typical limit of detection for a single measurement. Performing 5 min averages during these periods yield values <5 × 104 molecule cm−3.
Figure 7. Measurements from a nighttime flight off Midway Island. Takeoff for this flight was ∼1 hour after sunset. A relative altitude profile (0.5–6.7 km) is provided for reference. (a) Plot of observed [OH]. As can be seen, the values are quite small with typical 5 min averages <2 × 104 molecule cm−3. (b) Plot of observed [H2SO4] and [MSA] for the same flight. The boundary layer concentration of these two short-lived, photochemically produced species is quite large for nighttime conditions. As the typical lifetimes of these compounds are short (<20 min), these concentration levels indicate production is occurring. (c) Plot of [SO2] and [DMS], two reactants known to produce H2SO4 and MSA. The SO2 and H2SO4 measurements indicate that the oxidizing specie(s) is approximately as strong as OH (see text). Combining the very large boundary layer MSA concentrations with the rising concentrations, it is hard to explain that DMS is the primary reactant to form MSA.
Download figure to PowerPoint
 The noisy periods shown in Figure 7a are actually the indication of something quite unusual. The flight path was a straight line to the southwest, and a straight line return with boundary layer runs occurring at three fairly evenly spaced locations on the linear flight path. As can be seen from the altitude profile, the noisy periods occur during the portions of the flight flown in the boundary layer. Figure 7b is a plot of the H2SO4 and MSA concentrations from this same flight. As can be seen, both H2SO4 and MSA show a marked increase during these time periods, with H2SO4 reaching levels of 4–6 × 106 molecule cm−3 and MSA in excess of 107 molecule cm−3. The fact that such large concentrations of these two relatively short-lived, photochemically produced species are observed during the night is quite remarkable. Takeoff for this flight was ∼1 hour after local sunset. Typical lifetimes for H2SO4 and MSA in a clean environment are ∼20 min. The concentration levels of H2SO4 are on par with those seen during the day on the flights off Asia. Gas phase MSA concentrations such as these are typically only seen in much higher, dryer environments [Mauldin et al., 1999].
 These observations raise the concern that the large values could be due to evaporation of H2SO4 and MSA from the inlet surfaces. Large increases in temperature or drops in relative humidity could cause an increase in the vapor pressure of H2SO4 or MSA absorbed on the surface of the 1.9 cm curved inlet. The effect of evaporation has been observed by our group only once before [Mauldin et al., 1998], and then only at high altitude where the relative humidity fell below 5%. During the descents to these boundary layer runs, the temperature only changed by ∼30°C going from approximately −10 to +20°C. The relative humidity during these boundary layer runs never fell below 80%. From the work of Marti et al. , increasing the relative humidity from 5 to 80% decreases the vapor pressure of H2SO4 some 4 orders of magnitude. Field observations of MSA indicate that it has a similar but weaker response to relative humidity [Mauldin et al., 1998]. Thus both previous flight experience with the inlet and laboratory measurements of H2SO4 vapor pressure would indicate that evaporation of H2SO4 and MSA is insignificant for the conditions of the boundary layer portions (or the rest) of this flight.
 Evidence of nighttime oxidation has been seen before [Cantrell et al., 1997; Hu and Steadman, 1995; Berresheim et al., 2002]. Cantrell et al.  saw evidence of nighttime peroxy radical formation during the Mauna Loa Observatory Photochemistry Experiment (MLOPEX) 2c study. In that study the authors postulated NO3 radical chemistry as a possible explanation for their results. As these measurements were performed in the remote marine boundary layer, NOx/NOy levels were low. Measurements aboard the P-3B revealed average values of 0, 5, and 200 pptv for NO, NO2, and NOy, respectively during the boundary layer segments, indicating little possibility for NO3 formation. Additionally, rate coefficients show that NO3 will not significantly react with SO2 [DeMore et al., 1997]. Cantrell et al.  also suggested O3 chemistry as another possible source of OH and ultimately peroxy radicals. This chemistry involves the reaction of O3 with alkenes to forming a Criegee radical and ultimately OH. Berresheim et al.  have observed the presence of H2SO4 and MSA at in the coastal marine boundary layer at Mace Head, Ireland. The authors reported average nighttime concentrations of H2SO4 of ∼0.9 × 106 molecule cm−3 and performed experiments to show the presence of a specie(s) capable of oxidizing SO2 into H2SO4. A mechanism involving either Criegee radicals or BrO, or both, was postulated. It is doubtful that the Criegee chemistry can explain the results of the present study. The OH measurements during the boundary layer portions are scattered about zero and never rise above 2.2 × 105 molecule cm−3, a value far to low to maintain such high concentrations of H2SO4 and MSA. Additionally, as in the work of Cantrell et al. , the measured concentrations of O3 and alkenes during the boundary layer segments are too low to maintain the levels of H2SO4 and MSA observed. Data from the present study cannot confirm or rule out the possibility of oxidation involving halogen chemistry. The presence of MSA at night was explained by changes in mixing between the boundary layer and the mixed layer or free troposphere [Berresheim et al., 2002]. Such mixing allows MSA produced from DMS at higher altitudes to be entrained within the marine boundary layer [Davis et al., 1998; Jefferson et al., 1998]. This mixing is usually accompanied by the passage of a weather front. Examination of the meteorology for the study region, revealed no frontal passage for at least 36 hours prior to the flight. This fact together with the relatively short lifetime of gas phase MSA would tend to indicate that mixing of this type could not contribute significantly to the observed MSA concentrations.
 In order to maintain these daytime like concentrations of these two relatively short-lived species, production of H2SO4 and MSA must be occurring. In the marine environment these two compounds are primarily produced via oxidation of reduced sulfur, emitted from the oceans primarily in the form of DMS [Bates and Cline, 1985; Andreae et al., 1985]. Once in the atmosphere, DMS is oxidized by OH via a branched mechanism to produce either SO2, much of which eventually forms H2SO4, or products which can lead to the formation of MSA, [Ayers et al., 1991; Yin et al., 1990; Toon et al., 1987; Hatakeyama et al., 1982]. Figure 7c is a plot of SO2 and DMS concentrations also measured during this flight. As can be seen, both species exhibit concentration peaks during the boundary layer portions of the flight. The DMS concentrations are not large with boundary layer values increasing over time from ∼5 to ∼25 pptv. The SO2 concentrations are also modest with values of ∼50 pptv during all three boundary layer legs. The concentrations of both SO2 and DMS, while small, do indicate the presence of two known reactants that, when oxidized via OH, produce H2SO4 and MSA, respectively. However, the key oxidizer, OH, is not present, and this oxidation is occurring at night.
 An interesting exercise is to perform a crude “equivalent OH” behavior calculation for this “mystery” nighttime oxidizer. If a daytime scenario is assumed with production of H2SO4 from SO2 being in steady state equilibrium with its loss, an equivalent OH concentration can be obtained from the SO2 and H2SO4 measurements, assuming H2SO4 is formed from OH + SO2. Once formed, gas phase H2SO4 has an atmospheric lifetime on the order of an hour or less, being lost by condensation, rainout, or dry deposition. To properly calculate the H2SO4 lifetime, the particle surface area as a function of particle size needs to be known. Unfortunately, these values are not presently available, forcing the assumption of a H2SO4 lifetime. Here average respective SO2 and H2SO4 concentrations of 1.23 × 109 (50 pptv) and 5 × 106 molecule cm−3 and a rate coefficient of 8.83 × 10−13 cm2 molecule−2 s−1 for OH + SO2 + M H2SO4 + M [DeMore et al., 1997] were employed. H2SO4 lifetimes of <1 hour were assumed. Equivalent OH concentrations of 1.27–15.8 × 106 molecule cm−3 are obtained for H2SO4 lifetimes of 1 hour and 5 min, respectively, using this approach. While the equivalent OH value for a H2SO4 lifetime of 5 min (an extremely short lifetime) is large, lifetimes of 10 min to 1 hour yield values in the mid-106 molecule cm−3 range, on order with daytime measurements.
 Once formed in the gas phase, H2SO4 plays an important role in particle formation and growth. In areas away from the Earth's surface, condensation of H2SO4 is thought to occur predominately on the surface of pre-existing aerosol particles [Clarke et al., 1996; Ayers et al., 1991]. However, gas phase H2SO4 is thought to initiate new particle formation, if the existing particle surface area is small [Covert et al., 1992; Hoppel et al., 1994]. As previously mentioned, UCN was also measured during this flight. An analysis of this data during the boundary layer runs not only reveals the presence of larger more aged UCN with diameters <200 nm but also the presence of 3–4 nm and 3–8 nm UCN, indicating recent particle nucleation at night.
 As stated above, the MSA concentrations observed during the boundary layer runs are more typical of higher, dryer environments. In the remote marine environment, gas phase MSA is thought to be produced from the oxidation of DMS via OH. This oxidation can proceed via abstraction or addition channels, both of which produce MSA [Davis et al., 1998, 1999]. Other products include SO2 (which ultimately forms H2SO4), dimethyl sulfoxide (DMSO), dimethyl sulfone (DMSO2), and methane-sulfinic acid (MSIA). The overall temperature dependence of the reaction increases the final MSA/H2SO4 as the temperature decreases. Previous studies in areas with similar temperatures and much higher DMS production have yielded boundary layer daytime gas phase MSA concentrations on the order <106 molecule cm−3, levels substantially lower than those observed here in the dark [Mauldin et al., 1998, 1999, 2001]. Thus to maintain the concentration levels observed here, the production of gas phase MSA here at night must be greater than that from the OH + DMS reaction observed in those previous studies.
 Here two different possibilities can be considered: (1) the mechanism of DMS reacting with this nighttime oxidant is more efficient at producing MSA than the OH + DMS reaction, or (2) MSA is being produced via the oxidation of some other species. As the presence of this nighttime oxidant is only being inferred by the presence of short-lived oxidation products, any comment on a reaction mechanism with DMS would be speculation. However, there may be evidence that DMS is not the primary reactant. As can be seen from Figure 7c, the boundary layer concentration of DMS appears to be increasing over the evening, rising from 5 pptv on the first boundary layer leg, to 25 pptv on the last. The first and third legs were flown in the same area about 4 hours apart. This type of profile of DMS increasing over the evening has been observed before [Davis et al., 1999; Nowak et al., 2001], in an area with DMS production and little or no removal. Assuming DMS is the primary source of MSA, the rate of production (flux) DMS must be larger than the rate of removal by this nighttime oxidant for the same type of profile to be observed in the present study. This is a conclusion not out of the realm of possibility, however, it must be pointed out that it is a multistep reaction with OH that yields MSA from DMS, and presumably would be so with this nighttime oxidant.
 In the OH abstraction channel with DMS, MSA is produced via the formation of a CH3S radical followed by reactions involving O3, HO2, or NO [Davis et al., 1998]. The OH addition channel produces MSA via the formation of DMSO, followed by further reaction with OH. Comparing the rate coefficients for reaction with OH, one finds the DMSO rate coefficient is more than 100 times faster than that for DMS [DeMore et al., 1997; Hynes et al., 1986]. If the same holds factor holds true for the nighttime oxidant, it would take very little DMSO to produce the observed concentrations of MSA. The role of heterogeneous reactions could also be important in this oxidation. Jefferson et al.  postulate that the large particulate concentrations of MSA observed during the ground-based Sulfur Chemistry in the Antarctic Troposphere Experiment (SCATE) study on the Palmer peninsula, Antarctica, were due to condensed phase oxidation of DMSO. The high solubility of DMSO and the low vapor pressure of MSA would tend to preclude these sorts of reactions as being strong gas phase sources of either compound.
 Recently, Nowak et al.  have reported results from the GTE PEM-Tropics B study, indicating the possibility that DMSO may be produced at night. In that study the authors measured predawn boundary layer DMSO concentrations of 5–15 pptv. The DMSO concentration profiles were significantly out of phase with model predictions which assumed DMSO to be produced only from the OH initiated oxidation of DMS. In fact the measured diurnal DMSO profiles closely tracked the observed DMS profiles, increasing until the daylight hours and then decreasing. To reconcile differences between measurements and model predictions for the PEM-Tropics B study and to explain previous ground based DMSO observations, the authors point to an alternative DMSO source, possibly the ocean itself. The high solubility of DMSO in H2O, makes it difficult to understand how the ocean can be a very large source of gas phase DMSO, except possibly during large sea spray events. In light of the present results, another possible explanation of the observations of Nowak et al.  could be the formation of DMSO by the nighttime oxidation of DMS. There is some evidence of this possibility seen in the H2SO4 and MSA data obtained during the same PEM-Tropics B flight focused on by Nowak et al. . That flight was flown in the early morning such that the first boundary layer leg occurred around local sunrise. While not obtained in the dark, the H2SO4 and MSA concentrations were quite high for that time of day, with averages of ∼5 × 106 and ∼8 × 106 molecules cm−3, respectively [Mauldin et al., 2001]. Unfortunately, DMSO was not measured during the present study; thus its role in this nighttime oxidation chemistry remains speculative.