Model study of tropospheric trace species distributions during PEM-West A

. A three-dimensional mesoscale transport/photochemical model is used to study the transport and photochemical transformation of trace species over eastern Asia and western Pacific for the period from September 20 to October 6, 1991, of the Pacific Exploratory Mission-West A experiment. The influence of emissions from the continental boundary layer that was evident in the observed trace species distributions in the lower troposphere over the ocean is well simulated by the model. In the upper troposphere, species such as 03, NO v (total reactive nitrogen species), and SO2 which have a significant source in the stratosphere are also simulated well in the model, suggesting that the upper tropospheric abundances of these species are strongly influenced by stratospheric fluxes and upper tropospheric sources. In the case of SO2 the stratospheric flux is identified to be mostly from the Mount Pinatubo eruption. Concentrations in the upper troposphere for species such as CO and hydrocarbons, which are emitted in the continental boundary layer and have a sink in the troposphere, are significantly underestimated by the model. Two factors have been identified to contribute significantly to the underestimate: one is emissions upwind of the model domain (eastern Asia and western Pacific); the other is that vertical transport is underestimated in the model. Model results are also grouped by back trajectories to study the contrast between compositions of marine and continental air masses. The model-calculated altitude profiles of trace species in continental and marine air masses are found to be qualitatively consistent with observations. However, the difference in the median values of trace species between continental air and marine air is about twice as large for the observed values as for model results. This suggests that the model underestimates the outflow fluxes of trace species from the Asian continent and the Pacific rim countries to the ocean. Observed altitude profiles for species like CO and hydrocarbons show a negative gradient in continental air and a positive gradient in marine air. A mechanism which may be responsible for the altitude gradients is proposed. nisms convection


Introduction
As stated in the overview paper by Hoell  tures of model results and those of observations, particularly in the lower troposphere. On the other hand, because of the high wind velocity in the upper troposphere the model results tend to be strongly influenced by side boundary conditions [Brost, 1988;McKeen et al., 1991]. This problem can be alleviated by substantially increasing the model domain. However, relatively large computer resources are needed for these types of model calculations and will be a subject of future study.
The upper boundary of the model is located near the tropopause. Trace species such as O3 and NOy, which have a sig-nificant stratospheric source, are modeled by fixing their mixing ratios to values proportional to potential vorticity (PV) at the upper boundary [Ebel et al., 1991]. This allows us to study the budget of NOy in the upper troposphere, a subject of intensive investigation because of the potential impact of subsonic aircraft emissions on tropospheric O3 (e.g., NASA report by Prather et al. [1992]  water vapor mixing ratios are nudged (i.e., FDD-assimilated) throughout the model domain, as suggested by Stauffer and Seaman [1991]. Since the model (MM4) is very sensitive to the low-level temperature stratification, nudging of the temperature field is handled differently. Nudging the temperature within model-computed PBL height tends to affect the Blackadar planetary boundary layer (PBL) model drastically. In certain cases we find that the growth of PBL height is delayed by several hours in the morning. Therefore in this study the temperature nudging is only performed above 2 km. Below 2 km we allow the Blackadar PBL scheme to evaluate its own temperature profile. Finally, a parameterized vertical transport scheme due to convective clouds as described by Lin et al. [1994] is included in the model. In most cases, incorporation of FDDA into MM4 improves the model simulation significantly [Stauffer and Seaman, 1991]. For example, we have also made a simulation run of MM4 without FDDA for the PEM-West A period. In this run, neither of the two major typhoons, namely Mireille and Orchid, are fully developed. They stay as tropical depressions. Apparently, the default parameters used in the MM4 cumulus parameterization scheme (Kuo-type scheme) are not suitable for simulating the feedback between frictional convergence and release of latent heat inside the typhoon. The simulation is greatly improved when FDDA is incorporated. Figures la and lb depict the observed and model-simulated wind fields at the eleventh model level (1.15 km) at 1200 UT, October 4, 1991. The synoptic features are very well simulated, including a highpressure system over northern China (indicated by counterclockwise wind), and a quasi-stationary front is oriented from southwest to northeast from Taiwan to Japan (wind trough). In addition, both the position (18.8øN, 136.7øE) and strength of Typhoon Orchid are in good agreement with observations.

Emission Inventories
Anthropogenic emissions of NO•c and SO2 are based on the 1987 country or Chinese province level estimates of Karo and •lkimoto [1992]. The spatial partitioning within each country or province is taken from the 1 ø x 1 ø emission database of Dignon [1992] with modifications near coastlines to insure emissions are only over land in the 60 km x 60 km model grid. The only published estimates of anthropogenic hydrocarbons for Asia are those of Piccot et al. [1992], which are only available on a 10 ø x 10 ø or country resolution. These estimates do not include CO and are lumped into hydrocarbon classes that are not consistent with the Lurmann et al. [1986] oxidation mechanism. We have therefore recompiled our own estimates of hydrocar-

Initial and Boundary Conditions
The influence of side boundary conditions on a regional model has been well documented by Brost [1988] and by Mc-Keen et al. [1991]. In the boundary layer the distributions of trace species tend to be controlled by surface emissions and photochemistry within the model domain. At about 2 km the side boundary conditions start to affect significantly the distributions of trace species. Therefore unless the boundary conditions can be accurately defined, substantial errors can occur in the distributions of trace species calculated by the model. As Obviously, the initial and boundary conditions defined above are specified with certain degree of arbitrariness and contain significant uncertainties. As discussed above, when we compare the model results to observations, we will avoid any discussion of the altitude range (about 2 to 6 km) which is affected strongly by the side boundary or initial conditions.

Model Results and Comparison
With Observations

Horizontal Distributions of Trace Species
A straight forward display of the contrast between continental and marine air masses can be done by plotting the latitudelongitude distributions at various altitudes for the key species.
We will plot these distributions for 03, NO v, SO2, and CO and compare them to observations. These species are chosen because they play important roles in addressing PEM-West objectives. In addition, their distributions are affected by a wide variety of sources, sinks, and photochemical processes so that comparison of their distributions with obsewed values can be used as a test of the model's capability in simulating those atmospheric processes. For example, among the trace species measured during PEM-West A, CO turns out to be one of the most useful tracers for continental emissions not only because the majority of its sources are located in the continental boundary layer but also because of the large number of observations made by a fast response instrument [Hoe# et al., this issue].
When the model results are compared to observations, one important consideration is that while the mesoscale model is expected to simulate the gross features of large meteorological events reasonably well, the exact time and location of each event are not likely to be simulated accurately. We note that large features of the distributions of trace species calculated by the model during the study period do not drastically change with time. Therefore it is more meaningful to compare the time average distributions of the model results with observations rather than making comparisons of simultaneous model results with individual flights. The only obvious exception is the case when Typhoon Mireille brought relatively clean marine air up to the west of Japan and efficiently mixed boundary layer air to the upper troposphere. We will single out and discuss the effects of Typhoon Mireille on trace species distributions in a separate figure. A more detailed discussion of the typhoon and its effects will be presented in another paper by Newell et al.   The most obvious pattern in the model results is an increase of mixing ratio from the southeast to the northwest for 03, Nay, and sa2. This general pattern appears to be similar to that in the boundary layer. However, the cause of the pattern in the upper troposphere is completely different from that of the boundary layer. The former is due to the influence of a larger downward flux from the stratosphere in the northwestern region, whereas the latter is due to the influence of continental anthropogenic emissions that control the trace species distributions in the boundary layer.
Model-calculated distributions of 03 and sa2 are in good agreement with observed values both in the general pattern and in the absolute values. This is also the case for Nay, except that model-calculated values are about 30 to 60% lower than those observed. An exception to the general pattern is the relatively low mixing ratios in all three species observed during Typhoon Mireille, which is obviously due to transport of lowaltitude air by strong convective activities. This is also supported by relatively high mixing ratios of CO near southern Japan ( [Brost, 1988;McKeen et al., 1991] and transport from the stratosphere. Convective transport of trace species from the boundary layer to the upper troposphere can be substantial, but the model results indicate that significant convective transport occurs only in limited areas and/or special cases such as typhoons.

km (-221 mb
The above discussion suggests that underestimate of vertical transport and low mixing ratios specified for the upwind lateral boundary conditions (i.e., underestimate of the lateral transport from upwind regions) are two major reasons for the low mixing ratios of CO and hydrocarbons calculated by the model in the upper troposphere. They may also contribute to the low NOy mixing ratios calculated by the model. However, the underestimate of NO•. may also be due to the absence of lightning and aircraft sources in the model.

It should be noted that the upwind lateral boundary conditions are controlled by emissions upwind of the model domain. This implies that a significant amount of trace species observed in the free troposphere during PEM-West A came from regions outside of eastern Asia. Analysis of the atmospheric transport and photochemical characteristics of continental trace gases by Srnyth et al. [this issue] has reached a similar
conclusion.

An underestimate of the vertical transport from the boundary layer to the upper troposphere also implies an underestimate of the transport of upper tropospheric air to the boundary layer. As a result, the model tcnds to ovcrcstimate the contribution of in situ sources (including stratospheric intrusions) to the uppcr tropospheric distribution of trace species.
Prcvious studies [c.g., Startfief aml Sea,tan, 1991] indicate that the model makes an excellent simulation of the resolvable wind fields when FDDA is used (see also Figures la and lb), suggesting that resolvable vertical transport is well simulated. The underestimate in vertical transport by the model must be due to subgrid proccsses• particularly transport by convective activities. At this moment, without an extensive modeling study, wc are not able to evaluate quantitativcly the cll'cct of this model deficiency on the distributions of trace species in the upper troposphcre for the PEM-West A period. A preliminary analysis of the obscrvcd distribution of SO2 in the upper troposphcrc indicates the effect may be significant. This is probably a major source of uncertainty in this model study. It is important to qualify, however, that the above conclusion is drawn undcr the assumption that the vertical transport is accurately simulated in the model. The contribution of surface sources to the upper tropospheric NO r budget will increase with enhanced vcrtical transport. The model's underestimate of the vertical transport will lead to an underestimate of the importance of surface sources relative to upper tropospheric sources. Furthermore, the above study is limited to the upper troposphere over the western Pacific and is outside the tropical area. More observational and modeling studies are needed to evaluate the global budget of tropospheric NOr, particularly in the tropics.

Major observed characteristics of continental air were compared to those of marine air by Gregory et al. [this issue] and Talbot et al. [this issue]. In these studies, classification of continental and marine air masses was made by back trajectory analysis for air masses along the flight trajectories [Merrill, this issue
]. An air mass that passed over a continental area in the last five days or less was classified as continental air, while marine air was defined as an air mass that had not passed over any continental area in the last 10 days or longer. The authors compared the compositions between the two groups of air masses by plotting medians, quartiles, and ranges of trace spe-cies at various altitudes. Their plots for CO, C2H2, 03, and NO r are reproduced in Figures 2a-2d  To have enough data points, the time requirement of the marine air classification is reduced to 6 days the model results. This reduction in time requirement is not a problem, as it will become apparent later that the marine air defined above for the model has little contact with any continental air and has essentially the composition specified by the initial and boundary conditions over the ocean.
Another difference between the model's air mass classification and that of the observations is that the former determines back trajectories by following wind fields calculated by the model for grid points within the study area as indicated in Figure 1.

To preserve self-consistency in the model's results, it is necessary to use the model's wind fields instead of the isentropic back trajectories analyzed by Merrill [this issue]. We do not expect that the difference in air mass classification would lead to any significant bias in the model results because the study areas overlap substantially with the flight tracks and model wind fields have been shown to be in good agreement with results of isentropic trajectory analysis [Kuo et al., 1985]. As pointed out by Gregory et al. [this issue], difference in the altitude profiles between continental and marine air masses is an indicator of continental outflow for continentally emitted tracc species likc CO and C2H2
. However, the differencc alone does not give a measure of the flux of contincntal outflow. Additional information, namely, wind velocity and probability of occurrcncc of the continental air at each point, is needed to evaluate the flux. While the former can be obtained in the PEM-West data, the latter needs substantially more temporal and spatial coverage than is available from current observations. Nevertheless, comparisons of thc altitude profiles calculated by the model with observed valucs can provide a valuable test of how well the model simulates the transport processes. In the following, the comparison is done for CO and C2H2. Several conclusions can be drawn from the comparison and they apply to both CO and C2H:. For simplicity we use CO as the surrogate for discussion and make comments on C2H2 only when there is a significant difference.  ]. Thus the CO distribution above 9 km must be controlled by long-range transport processes with scales greater than the model domain. However, this does not imply that globally the vertical flux from lower troposphere to 9 km and above is smaller than at lower altitudes. The flux can be larger, but horizontal transport and dispersion processes above 9 km are so fast that CO transported from below is mixed quickly and thoroughly so that little difference can be observed in the CO mixing ratio between continental and marine air masses.
Finally, in the marine air there is a positive altitude gradient in the observed CO mixing ratio. It is most evident when the highest layer (above 9 km) is compared to the lowest three layers (below 3 km). The positive altitude gradient exists also in the altitude profiles of O3 and NOy. In fact, the positive gradient in altitude profile is present in all trace species which are emitted over the continent. What is the cause(s) of the positive altitude gradient in the mixing ratio of a trace species in the marine air? Obviously, a stratospheric or upper tropospheric source can result in a positive gradient. This is evident in the positive gradient in marine air simulated by the model for O3 and NOy. For other species, a different process(es) must be responsible for the positive gradient. In a recent study of altitude distribution of nonmethane hydrocarbons, Blake et al. [1992] found similar positive gradients and suggested three different types of meteorological conditions that might contribute to the positive gradient in their data. All three included an efficient vertical convection process over continental source areas. Their differences were in the detailed transport mechanisms of the convection and what followed the convection. The first included a thoroughly mixed column of air at all altitude levels plus a greater photochemical sink in the lower altitude due to a higher OH concentration. The second required a faster horizontal transport in the upper troposphere than the lower troposphere so that the upper tropospheric air would be exposed to less reaction with OH. The third required a cloud pumping mechanism like the one proposed by Gidel [1983] that rapidly pumped boundary layer species up to and deposited them in the upper troposphere. Blake et al. [1992] did not indicate any preference but suggested a mixture of them. PEM-West A data have shed more light to this question. The fact that there is a negative altitude gradient for CO and C2H2 mixing ratios in the continental air (air over the ocean with continental origin) despite the higher OH concentration and slower wind velocity in the lower troposphere suggests that the third type is not the dominant mechanism because it would result in a positive vertical gradient for the continental air. Vertical transport or mixing from the surface to the lower and middle troposphere needs to be efficient enough to maintain the negative gradient in the continental air. However, because of the tendency of a faster horizontal dispersion of trace species in the upper troposphere due to stronger winds, we cannot determine the relative magnitude of vertical transport to various altitudes.
A mechanism that can explain the altitude gradients of CO and C2H 2 for both continental air and marine air may be described as follows. When air masses pass over continental areas, they entrain boundary layer emissions through convection or other vertical transport processes. The entrainment as well as detrainment need to be significant at all altitudes. Furthermore, to obtain a negative gradient for the entrained trace species in the continental air, the entrainment must be large enough at lower altitudes to overcome the effect of a greater sink due to greater OH and smaller wind velocity. In other words, direct cloud pumping to the upper troposphere alone would not work. For the marine air, the positive gradient of trace species can be maintained by a combination of the following two processes: first, fast dispersion of CO and C2H2 transported to upper troposphere due to strong horizontal wind and turbulence, and second, greater OH concentration and lower wind velocity in the lower troposphere which results in air masses staying longer over the ocean in the lower troposphere than in the upper troposphere.
Observed NOy and 03 altitude profiles look similar to those of CO and C2H 2 except that for continental air the negative altitude gradient found in CO and C2H: is absent in NO• and more so in 0 3. In fact, the latter has a positive gradient. These features which are obviously due to stratospheric and/or upper tropospheric sources are simulated very well by the model, particularly for 03. The model underestimates NOy in the marine air by about 50%. As discussed in sections 2 and 3, this may likely be due to the omission of important free tropospheric sources such as lightning and aircraft emissions.

Summary and Conclusions
A three-dimensional mesoscale transport/photochemical model is used to study the transport and photochemical transformation of trace species over eastern Asia and the western Pacific for the period from September 20 to October 6, 1991, of PEM-West. Latitude-longitude distributions for a set of key species in the boundary layer and in the upper troposphere from model results have been compared to aircraft measurements. The influence of emissions from continental boundary layer that is evident in the observed trace species distributions within the boundary layer is reasonably well simulated by the model.
In the upper troposphere, species which have a significant source in the stratosphere are simulated well in the model. These species include 03, NOy, and SO2, suggesting that the upper tropospheric abundance of these species are controlled to a large extent by stratospheric sources. In the case of SO2 the stratospheric source is identified to be mostly from the Mount Pinatubo eruptions which occurred about three months before the PEM-West A mission.
Concentrations in the upper troposphere for species such as CO and hydrocarbons which are emitted in the continental boundary layer and have a tropospheric sink are significantly underestimated by the model, indicating that vertical transport processes including convection may be underestimated in the model. In addition and probably more important, low mixing ratios specified for the boundary conditions which are controlled by sources upwind of the model domain also contribute to the underestimate by the model for species like CO in the upper troposphere. In the future, hemispheric scale model calculations will be carried out to study this question. We have made an analysis of the contribution of various sources to the upper tropospheric NOy over the western Pacific at midlatitudes. Assuming that the vertical transport is accurately simulated in the model, we estimate that stratospheric N20 oxidation, subsonic aircraft, and the rest (lightning plus surface sources) contribute approximately 25, 50, and 25%, respectively. However, there is evidence indicating an underestimate of the vertical transport by the model that will lead to Observed altitude profiles for species like CO and hydrocarbons show an interesting difference between continental and marine air masses, namely, a negative gradient in continental air and a positive gradient in marine air. A mechanism that can explain the altitude gradients has been proposed as follows: entrainment of boundary layer trace species at all altitude over the continent through convection and other vertical transport processes, fast dispersion of the trace species by strong horizontal wind and turbulence in the upper troposphere, greater OH concentration at lower altitude, and lower wind velocity in the lower troposphere which leads air masses to stay longer over the ocean than those in the upper troposphere.