Trace gas transport and scavenging in PEM-Tropics B South Pacific Convergence Zone convection

. Analysis of chemical transport on Flight 10 of the 1999 Pacific Exploratory Mission (PEM) Tropics B mission clarifies the role of the South Pacific Convergence Zone (SPCZ) in establishing ozone and other trace gas distributions in the southwestern tropical Pacific. The SPCZ is found to be a barrier to mixing in the lower troposphere but a mechanism for convective mixing of tropical boundary layer air from northeast of the SPCZ with upper tropospheric air arriving from the west. A two-dimensional cloud-resolving model is used to quantify three critical processes in global and regional transport: convective mixing, lightning NO x production, and wet scavenging of soluble species. Very low NO and 03 tropical boundary layer air from the northeastern side of the SPCZ entered the convective updrafts and was transported to the upper troposphere where it mixed with subtropical upper tropospheric air containing much larger NO and 03 mixing ratios that had arrived from Australia. Aircraft observations show that very little NO appears to have been produced by electrical discharges within the SPCZ convection. We estimate that at least 90% of the HNO 3 and H20 2 that would have been in upper tropospheric cloud outflow had been removed during transport through the cloud. Lesser percentages are estimated for less soluble species (e.g., <50% for CH3OOH ). Net ozone production rates were decreased in the upper troposphere by -60% due to the upward transport and outflow of low-NO air on of SPCZ.


Introduction
The South Pacific Convergence Zone (SPCZ) is one of the Earth's most expansive and persistent convective cloud bands and plays an important role in global circulation patterns. Although this meteorological feature was first noted in the 1930s, it was not until Trenberth [1976]   Deep convection has been shown in numerous studies [e.g., Dickerson et al., 1987;Pickering et al., 1990Pickering et al., , 1996] to be effective at rapidly transporting boundary layer air to the upper troposphere. As a result, free tropospheric ozone production is perturbed due to the convectively induced changes in NOx and HOx. Over the remote oceans, marine boundary layer air with low NOx and low 03 mixing ratios is transported upward by deep convection [e.g., Pickering et al., 1993] tending to decrease the magnitude of ozone production in the upper troposphere. LelieveM and Crutzen [1994] suggested that convection over the remote oceans tends to bring higher values of NOx and 03 down to the boundary layer where their lifetimes are relatively short, thereby causing a decrease in tropospheric ozone. Although lightning is much less frequent in marine convection than in continental storms [e.g., Christian and Latham, 1998], significant perturbations to NOx have been noted in association with marine storms [e.g., Chameides et al., 1987]. The large amounts of liquid water in tropical convection may efficiently remove soluble gases and aerosols from the atmosphere. The results of transport, lightning, and scavenging analyses for the observed SPCZ convection will increase our understanding of these processes over the tropical oceans and will contribute to improvements in parameterizations of them in global chemical transport models. Section 2 of the paper provides background material on the climatology of the SPCZ and the meteorology of the SPCZ for the PEM-Tropics B convective event that was sampled by the DC-8 and simulated with the cloud model. Section 3 presents the DC-8 meteorological and chemical measurements from the SPCZ flight, and section 4 details our cloud model analysis. In section 5 we discuss our results in relation to previous work, and section 6 summarizes the findings of our analyses.

Climatology
The SPCZ is a zone of primarily low-level convergence consisting of a zonal portion to the northwest of New Guinea where it joins with the Intertropical Convergence Zone (ITCZ) and a diagonal portion which extends southeastward from New Guinea to approximately 30øS and 150øW (see schematic in Figure 1). It is most well-defined in January and February; therefore the PEM-Tropics B field mission of March and early April occurred after the typical time of maximum intensity. A thorough review of the characteristics of the SPCZ, its origin and maintenance, and its significance in the global circulation has been published by Vincent [1994]. Streten [1973] found through analysis of satellite imagery that the SPCZ contained a higher percentage cloud cover than any other region of the Southern Hemisphere. These values maximize in the Southern Hemisphere summer, along with the synoptic-scale vertical velocity. At this time of year there is also a minimum in outgoing longwave radiation in the SPCZ. Kiladis et al. [1989] have shown that maximum surface wind convergence in the SPCZ occurs south of the axis of maximum precipitation and the maximum precipitation lies south of the region of maximum sea surface temperature. Huang and Vincent [1983] showed that the sea level pressure trough associated with the SPCZ lies along the southwest edge of the SPCZ cloud band. This arrangement suggests that the deep convection forms primarily in the tropical air to the north of the convergence zone. We employ this concept in section 4 when we select a profile for initializing the cloud model simulation.
The zonal portion of the SPCZ may be maintained due to the "warm pool" of seawater in this region. The sea surface temperature gradients may force the low-level winds [e.g., Lindzen and Nigam, 1987] resulting in convergence. The diagonal portion of the SPCZ, which is the portion near Fiji, results from a tropical-midlatitude interaction. The convergence zone results from the northeasterly flow on the northwest side of the subtropical eastern Pacific high and the slightly cooler and drier southeasterly flow from the subtropical latitudes of the Southern Hemisphere located farther to the south and west. This convergence zone interacts with transient troughs in the midlatitude westerlies. The frontal activity associated with these troughs strengthens the convergence between them and the subtropical high. The fronts in the westerlies propagate equatorward from near New Zealand and often become quasistationary in the SPCZ.
Fuelberg et al. [this issue] summarized positions of the SPCZ for each PEM-Tropics B flight. Greater variability in the position was noted in the far southeast portion of the SPCZ due to the influence of midlatitude cyclones and cold fronts. The fronts moved to the northeast and became stationary at 15 ø-20øS. The convergence associated with the front became connected with the SPCZ. Figure 2 shows  negative anomalies (implying more frequent or deeper convection than normal) existed at the western end of the SPCZ and strong positive anomalies (implying less convection than normal) are seen to the east, the anomaly was near zero in the vicinity of Fiji. Therefore it appears that the strength of the SPCZ convection near Fiji in March 1999 was near normal.
The anomalies noted are related to the La Nifia conditions that existed in 1999. While the SPCZ convection near Fiji appeared near normal, the transport characteristics of the region may not have been normal due to the strong anomalies in convection to the east and west. 1999 On March 21, 1999, the SPCZ was located just north of Fiji as seen in Plate 1, which is an infrared cloud image from the GMS satellite at 0232 UT. The flight track of the DC-8 aircraft is superimposed on the image. After takeoff from Fiji, the aircraft sampled air on the southwest side of the SPCZ. It then passed through the SPCZ convective cloud band at 10.7 km and continued to the northeast where weaker convection was sampled at 8.8 to 11.3 km. The plane turned around at ---10øS and descended through the weaker cloud band, reaching the boundary layer on the southwest side. The marine boundary layer on the northeast side of the SPCZ was sampled for about 20 min prior to an ascending spiral to 11.3 km. The aircraft then passed through the SPCZ at this altitude, followed by a descending spiral to the boundary layer on the southwest side of the SPCZ convection prior to the return to Fiji.

Meteorology of SPCZ Convection of March 20-21,
Plate 2 shows the evolution of the minimum cloud top pres-  convergence was strongest along a line from --• 13øS, 170øE to -20øS, 170øW. At low levels, air approached the northeast side of the SPCZ near Fiji after a long transit across the Pacific at tropical latitudes of the Southern Hemisphere. On the southwest side, air approached from the southeast. Therefore it appears that in the vicinity of Fiji the SPCZ represented a convergence zone for two Southern Hemispheric air masses. However, well to the northwest of Fiji the tropical air approaching the SPCZ had a Northern Hemispheric origin. There the SPCZ is a line of convergence of air masses from the two hemispheres. At 500 hPa there were easterly winds on the northeast side of the SPCZ and northwesterly winds on the southwest side. At higher levels the westerlies were relatively strong southwest of the SPCZ, while easterlies still prevailed to the northeast of the SPCZ.

Measurements
In this section we describe the set of measurements taken by some of the smallest measured during PEM-Tropics B and result from the long fetch of the northeasterly winds with ozone being destroyed at the ocean surface during transport. In contrast, boundary layer O3 on the southwest side was 17-19 ppbv with lower free troposphere values of 20-35 ppbv. In addition, two pronounced O3 plumes were found on the southwest side. The upper plume with maximum ozone of --•60 ppbv was located in the 10-11.3 km layer. The lower plume was in the 6.5-8 km layer and also reached a maximum of --•60 ppbv.  Figure 6 shows the NOx and HNO3 profiles on both sides of the SPCZ. The northeast side of the SPCZ was characterized by very small mixing ratios of NOx (Figure 6a). On the southwest side a NOx plume of --•50 to 75 pptv in the same layer as the upper ozone plume may signify that lightning NOx catalyzed ozone production during the 3 days transport from Australia. Little nitric acid ( Figure 6d) and hydrogen peroxide were found in this plume, supporting an interpretation of convective outflow following scavenging. In the lower altitude ozone plume (--•7 km altitude) there was little NOx, but substantial plumes of nitric acid (up to --•275 pptv), H202, and PAN were observed. This could suggest a nonconvective origin (no wet scavenging) or that the source of the plume was located further back in time than for the upper plume, allowing more conversion of NOx to nitric acid and PAN. In addition, a small enhancement in CO (---6 ppbv) was also evident. Figure  7 shows a cluster of 10-day back trajectories for the lower altitude ozone plume. The majority of these trajectories passed over southern Australia 4 to 5 days back in time. A frontal system with moderate cloud tops (7-9 km) crossed southern Australia during these 2 days. Polluted air may have been lofted ahead of the cold front followed by photochemical processing during transport. Figure 8 shows very sharp gradients of CH3I, a tracer of marine origins [Davis et al., 1996a;Cohan et al., 1999] within the boundary layer. On the northeast side of the SPCZ, mixing ratios decreased with altitude from 0.47 pptv at 0.4 km to <0.2 pptv at 1 km. Boundary layer values were larger on the southwest side of the SPCZ, decreasing with altitude from 0.72 pptv at 0.3 km. Mixing ratios remained >0.2 pptv at all altitudes below 2 km. These boundary layer maxima allow CH31 to be an excellent tracer of marine convective transport from the boundary layer to the free troposphere as demonstrated in the following subsection. During the southbound (second) SPCZ crossing, sudden decreases of ozone and CO to minima of 10 ppbv and 40-42 ppbv, respectively, were noted. It appears that little mixing of the cloud outflow with the upper tropospheric air arriving from the east had occurred on the northeastern edge of the cloud band. In this case the aircraft passed just to the west of the dominant cell. Therefore it is not surprising that more pronounced minima in ozone and CO were noted (0320-0324 UT) than during the first SPCZ crossing. Following these minima, both CO and 0 3 gradually increased as the DC-8 approached the southwestern edge of the convective band. At 0335 UT the winds shifted from easterly to westerly near the edge of the cloud, and by 0350 UT the air with much larger ozone mixing ratios on the southwest side of the SPCZ was encountered. Therefore the region between the maximum convective outflow (---0324 UT) and the subtropical air (---0350 UT) represents a zone of mixing of tropical marine boundary layer air with air in the subtropical upper troposphere. Figure 10 presents the observations of CH3I taken during the SPCZ traverses. Maximum values of CH3I mixing ratio correspond closely in time with the ozone minima. These maxima reached 0.28 pptv during the northbound crossing and 0.36 to 0.40 pptv in the more intense convective outflow encountered during the southbound crossing. Minimum values of 03 and maximum values of CH3I noted during the SPCZ crossings suggest that most of the air entrained into the SPCZ convection derived from the northeast side of the convective system. If substantial boundary layer air from the southwest side of the SPCZ had entered the system, the pronounced ozone minima would not have been found in the convective outflow, and the CH3I values would have been larger. This result is in agreement with climatological analyses (see section 2.1) which indicate that the SPCZ convection typically develops in the tropical air on the northern edge of the zone of convergence. Figure 11 shows the mixing ratios of NOx and HNO3 during the SPCZ crossings. NO x mixing ratios dropped to as low as 10 pptv during the southbound traverse. HNO3, which is an extremely soluble gas, was substantially depleted, especially during the southbound crossing. Values as low as 5 pptv of HNO3 were measured.

Model Analyses
The observations discussed in section 3 provide information on the net effect of several processes associated with convective clouds. In combination with a cloud model, the observations can yield further information concerning the individual processes of convective transport, wet scavenging of soluble species, and production of NO x by lightning.

Ensemble (GCE) Model
The GCE model has been used extensively for studies of convective transport of trace gases in both the tropics and midlatitudes [e.g., Scala et al., 1990;Pickering et al., 1991Pickering et al., , 1992Pickering et al., , 1996Pickering et al., , 1998Stenchikov et al., 1996;DeCaria et al., 2000]. Here we use the model to investigate trace gas transport in a representative convective system in the SPCZ near Fiji.
A complete description of the GCE model and its governing equations is given by Tao and Simpson [1993]. The model uses a parameterized Kessler-type liquid water scheme [Kessler, 1969] for cloud and rainwater and a parameterized ice phase scheme for cloud ice, snow, and hail/gaupel [Lin et al., 1983]. The two-dimensional (2-D) version of the model in use at the University of Maryland does not include radiative transfer processes because they are not significant on the several-hour timescale of the simulations conducted in this study .  Table 1 for the nine tracer layers of the model. The first three of these species are insoluble, with lifetimes considerably longer than a convective system, and can be used as tracers of convective transport and mixing. We simulate the transport of NOx, primarily to enable the estimation of the contribution of lightning to the observed NOx mixing ratios (see section 4.4.2). The latter five species are soluble, and we perform conserved tracer transport simulations to estimate the fraction of amounts of these species that are scavenged in the cloud. Figures 15a and 15b present results for the ozone tracer at 2 and 4.5 hours in the simulation. Initially, ozone in the model at the DC-8 flight level during the SPCZ crossings (---11.3 km) was ---28 ppbv. The simulated convection transported lower values from the marine boundary layer to the middle and upper troposphere, so that 11-km mixing ratios decreased to ---11.5-13 ppbv. These values compare very well with the minimum (12-13 ppbv) recorded on the DC-8 during the northbound SPCZ crossing (Figure 9a). The model slightly overestimates the minimum ozone mixing ratio (10 ppbv) recorded on the southbound crossing. Ozone averaged over the cloudperturbed region in the simulation was 15 ppbv at 2 hours and 17 ppbv at 4.5 hours, slightly overestimating the observed averages of 14.5 ppbv and 14 ppbv on the northbound and southbound crossings, respectively. The effect of the convection on  Figures 17a and 17b present tracer transport results for NOx with NOx treated as a passive tracer with no sources or chemistry. Chemical loss is expected to be small over the short timescale of convective transport. However, in reality, NOx could have been produced by lightning. At cloud outflow levels the model calculation represents NOx mixing ratios with no production by lightning. Therefore we compare these values with those observed during the SPCZ crossing in order to make an estimate of the possible effect of lightning NOx production by examining the difference between the two quantities. Observed NOx (actually a box model estimate based on the observed NO) during the northbound SPCZ crossing was 16 to 20 pptv, while NOx at flight level from the tracer transport calculations was --•5 pptv. Therefore the difference, 11 to 15 pptv, could be attributed to lightning NOx production. This is a very small quantity, suggesting that the SPCZ cloud was not very electrically active. Unfortunately, the   mixing ratios in the outflow (i.e., at or below the instrument detection limit). Somewhat smaller fractions scavenged (---0.7) are estimated for the outer part of the cloud anvil. The cloud model output at ---11 km indicates that the hydrometeors at this altitude were almost all in the form of snow, graupel, or cloud ice. Therefore these results suggest that HNO 3 either was retained by the hydrometeors during the freezing process or uptake onto the ice particles occurred. Similar calculations were performed for H202, CH3OOH, H2CO , and SO2 (see Table 2). A large fraction scavenged (0.91) was obtained for H202, which is also very soluble. We obtain a considerably lesser fraction for CH3OOH (   The lowest ozone mixing ratios computed in the upper troposphere were ---2 ppbv, an 80% decrease from the preconvective conditions. This ozone mixing ratio is considerably smaller than the minimum value found in the SPCZ case analyzed here (---10 ppbv). In another CEPEX event, Wang et al. [1995] simulated tracer transport and estimated a minimum ozone value of 9 ppbv in the cloud anvil, which closely compares with that observed in the SPCZ.
We noted only very small enhancements of upper tropospheric NOx possibly due to lightning during one of the SPCZ crossings. However, the Stormscope on board the DC-8 recorded an average of 2-3 discharges per minute as the plane approached and passed through the convective system. Pickering et al. [1998] simulated lightning NO production in a squall line that occurred during TOGA-COARE near the Solomon Islands in the tropical Pacific. When the model parameterization was adjusted to produce the measured flash rates of 1-2 per minute, it yielded maximum NOx mixing ratios in the anvil of 0.7-0.9 ppbv. These values are similar to those reported by  likely that most of the SPCZ Stormscope signals were from discharges not sufficiently energetic to produce significant quantities of NO.

Summary
We have analyzed and simulated with a cloud-resolved model trace gas transport and scavenging in SPCZ deep convection observed during the PEM-Tropics B experiment. We have examined in detail the structure of convective mixing, the production of lightning NO, and the wet scavenging of several soluble species, which are three of the most uncertain processes in regional and global chemical transport models (CTMs). Wind fields from a 2-D cloud-resolving model were used with tracers to diagnose transport characteristics, lightning NO production, and wet scavenging. The conclusions are as follows: 1. For transport, from the model and the observations, we conclude that air in deep convective outflow sampled by the DC-8 was mostly derived from the marine boundary layer in the tropical air mass northeast of the SPCZ. It appeared that very little boundary layer air from southwest of the SPCZ was entrained into the convection. Therefore, at low levels, the SPCZ acts as a barrier to mixing. However, because the SPCZ convection changes position from day to day and is more active on some days than others, this low-level barrier is spatially and temporally transient. The observations suggest that most of the mixing of the tropical air from northeast of the SPCZ with Southern Hemispheric subtropical air from southwest of the SPCZ occurred after the tropical air had been processed by the convective cloud. This mixing took place in the upper troposphere on the southwest edge of the convective cloud band. At the portion of SPCZ northwest of Fiji the tropical air north of the SPCZ had a Northern Hemispheric origin. Therefore, during PEM-Tropics B, it is likely that this part of the SPCZ served to mix air from the two hemispheres.
2. For lightning, by comparing model-calculated tracer transport of NOs to observed outflow levels, we conclude that very little NO appears to have been produced by lightning within the SPCZ convection, despite vertical velocities that were fairly strong for a tropical marine system. However, during the northbound passage the aircraft was not at the best location to capture maximum outflow, and NO data were not collected during most of the southbound passage. 3 SO2, and H2CO ). Convective transport of low-NO marine boundary layer air to the upper troposphere reduced net ozone production in this region by -60%. This air mixed with higher-NO air arriving from the west at these altitudes, resulting in air parcels with ozone production rates intermediate between the preconvective rates on either side of the SPCZ. Therefore the SPCZ modifies the ozoneproducing capacity of the South Pacific troposphere. Comparison of parameterized convective mixing, lightning NO production, and wet scavenging from regional and global model simulations of the March 20-21, 1999, SPCZ event with the observational and cloud-resolving model results presented here serves as a means of evaluating and refining such parameterizations. Therefore, if parameterization improvements result from the comparisons, detailed process analyses for specific events such as that presented here will likely aid in reducing uncertainties and inaccuracies in regional and global CTMs.