Vertical transport of tropospheric aerosols as indicated by 7Be and 210Pb in a chemical tracer model

We use the natural radionuclides 7Be and 210Pb as aerosol tracers in a three‐dimensional chemical tracer model (based on the Goddard Institute for Space Studies general circulation model (GCM) 2) in order to study aerosol transport and removal in the troposphere. Beryllium 7, produced in the upper troposphere and stratosphere by cosmic rays, and 210Pb, a decay product of soil‐derived 222Rn, are tracers of upper and lower tropospheric aerosols, respectively. Their source regions make them particularly suitable for the study of vertical transport processes. Both tracers are removed from the troposphere primarily by precipitation and are useful for testing scavenging parameterizations. In particular, model convection must properly transport and scavenge both ascending 210Pb and descending 7Be. The ratio 7Be/210Pb cancels most model errors associated with precipitation and serves as an indicator of vertical transport. We show that over land the annual average 7Be/210Pb ratio for surface concentrations and deposition fluxes vary little globally. In contrast, the seasonal variability of the 7Be/210Pb concentration ratio over continents is quite large; the ratio peaks in summer when convective activity is maximum. The model overestimates 7Be in the tropics, a problem which we relate to flaws in the GCM parameterization of wet convection (excessive convective mass fluxes and no allowance for entrainment). The residence time of tropospheric 7Be calculated by the model is 23 days, in contrast with a value of about 9 days calculated for 210Pb, reflecting the high‐altitude versus low‐altitude source regions of these two tracers.


tive. For example, convection pumps •Rn and 2•øPb upward and carries •Be downward. Dry convection is
most active during the summer, producing summertime maxima in the 7Be/•løPb concentration ratio over continents. Convective scavenging, which is active in the tropics year-round and at higher latitudes during the summer over continents, produces asymmetries in removal since clouds tend to form over rising air leading to scavenging of upward moving species [Giorgi and Chainaides, 1986;Rodhe, 1983]. B93 developed a parameterization in the CTM for scavenging aerosols in precipitating convective updrafts. This parameterization performs reasonably well for •øPb, which has relatively high concentrations near the surface. However, we will show that ?Be concentrations are overpredicted by the model in regions where convection is active. This is a consequence of some combination of excessive downward transport and insufficient scavenging in convection. Thus the difference in the response of •Be and •XøPb to convective transport and scavenging provides 1985]. Indeed Feely et el., [1988] emphasized that surface concentrations are influenced to varying degrees at different locations by stratosphere-troposphere exchange, precipitation scavenging, vertical transport in the troposphere, horizontal transport, and radioactive decay. We use the CTM, together with the information obtained from simultaneously modeling 2XøPb, to help distinguish between the importance of these factors on ?Be concentrations and deposition at different locations. In particular, our study confirms the prominent role of convection in determining ?Be (and 2XøPb) seasonality at midlatitude continental sites, as argued by Feely et el., [1988].
Beryllium 7 and •løPb were modeled by Brost et el., [1991] and Feichter et el., [1991], respectively, using the three-dimensional tracer model ECHAM2, which was developed from the European Centre for Medium Range Weather Forecasting (ECMWF) model. These studies used a first-order scavenging parameterization [Giorgi and Chainaides, 1986] for both convective precipitation and large-scale precipitation. The lack of precipitation scavenging within convective cloud updrafts resulted in 2•øPb surface concentrations which were at least 40% higher than measured values, while deposition fluxes were underestimated. Agreement for •Be was much better. They did not make extensive use of the combination of 2xøPb and •Be, although they commented on the utility of such a study. Rehfeld and Heimann [1995] used the ECMWF model to simulate 2xøPb, 7Be, •øBe, and 9øSt. They also used a first-order scavenging parameterization but were more successful than Fetchtar et el. [1991] at simulating 21øPb surface concentrations (the root mean square value between observations and model was about 0.2 mBq/m a STP [Rehfeld, 1994]). Rehfeld and Helmann [1995] incorporates observed meteorological winds and precipitation from 1990, and they compared model results with •Be and 2XøPb data collected in 1990. In contrast, we use i year's output from the GISS GCM 2, which generates internally consistent climatology, and we compare results with observations averaged over all available years.

Beryllium 7 Source
We will confine our discussion to the •Be source, since the •øPb source is described by B93. Beryllium 7 is generated by cosmic rays, which travel along magnetic field lines, so that the source is strongest near the poles (see Figure 1). The cosmic rays collide with atoms in the upper atmosphere, generating a cascade of neutrons and protons, which in turn interact with nitrogen and oxygen atoms, resulting in •Be production. The encounter of the neutrons and protons with the nitrogen and oxygen atoms is called spelletlon or "star" production. The magnitude of •Be production is a trade-off between the energy level of the bombarding particles and the density of the atmospheric N and O targets, so that maximum production occurs in the stratosphere ( Figure 1). Very soon after production, 7Be attaches to available aerosol particles. •Be production has neglible dependence on season or longitude.
In addition to the dependence of production on altitude and latitude, production also varies with the • 11-year solar cycle. When solar activity is high, cosmic rays are deflected away from the solar system, and •Be production decreases. Figure 2 shows  ?Be concentrations at four sites with long data records.   The 7Be source was constructed by fitting cubic splines to production values provided by D. Lal (personal communication, 1994) (shown in Figure 1) and computing the average VBe production within each model layer as a function of latitude. Because of the large changes in production with altitude we calculated and used firstorder vertical moments of the source as described by Prather [1986].
The GCM and CTM distinguish between large-scale precipitation, shallow wet convection (events confined below layer 4), deep wet convection, and dry convection. The vertical distribution of convective events and large-scale precipitation is archived from the GCM as 5-day averages, so that event heights are reconstructed in the CTM by combining this information with the 4-hour archives of event occurrences. For dry convec-five events, when air is unstable with respect to the dry adiabat, the unstable column is mixed uniformly. For wet convective events, when air within the column is unstable with respect to the wet adiabat, 50% of the air in the lowest grid box is moved upward to the point where its density is in equilibrium with the environment and is deposited in that layer, followed by subsidence of dry air within intermediate layers as required to conserve mass. The scheme then checks each layer above for instability; however, 50% mass flux is generally far greater than is required to stabilize the column. This GCM version did not include cloud entrainment, anvil precipitation, or cloud downdrafts. B93 employed a wet convection scavenging parameterization consistent with the vertical mass flux occurring in the GCM convective events. A given percentage of aerosol is removed from the upward moving air for all wet convective events. B93 used 100% removal for deep convective events and 50% removal for shallow convection. While this procedure was reasonably successful for simulating •'•øPb, we found that it resulted in excessive VBe surface concentrations in the tropics and in the summers at midlatitudes. One reason for this is that scavenging only occurs within air moving upward in the updraft, i.e., in air originating in the lowest convective layer, since convection in this GCM is nonentraining.
Hence this mechanism efficiently removes •'•øPb relative to 7Be because a much larger percentage of the •'•øPb inventory exists in the lower troposphere. The model also does not account for scavenging in the cloud anvil, or the laterally sheared cloud top. To compensate for these model deficiencies, we applied a first-order rainout parameterization to regions of convective precipitation in addition to the convective scavenging parameterization of B93. We will show that this added scavenging improves the simulation in the tropics for both VBe and •-•0pb ' The Giorgi and Chaincities [1986] first-order rainout parameterization is used for both large-scale and convective precipitation, in addition to the scavenging in wet convective updrafts described by B93. The fraction of the grid box experiencing precipitation is given by 1• = roqat/T: where Q is the model-derived mean water condensation rate, Zkt -4 hours is the model Since most 7Be production occurs at high altitudes, we use stratospheric measurements to verify the source. Although the data are sparse, stratospheric concentrations should remain relatively constant. In Figure 4 we compare model-generated annual average VBe con- The GCM is intended to simulate a typical meteorological year, rather than a specific year; evaluation of model results with observations must therefore focus on long-term statistics rather than on measurements for any specific day.   [Feel• et al., 1981, 1988, Larsen and Sanderson, 1990, 1991 (Table 1). Increasing convective scavenging fur-tions (mBq/m 3 STP) as a function of latitude.  ously discussed GCM flaws in simufating tropical con-ratio) and Figure 12 (the surface concentration) has a vection. Both •Be a/ad the •Be/•'mPb ratio are high. slight positive poleward gradient, so that the deposition The GCM also does not accurately simulate the sea-flux ratio increases toward the poles faster than does the sonal motion of the intertropical convergence zone, which surface concentration ratio. The reason for this is shown is shifted too far north during the summer. Thus the seasonality of rainfall in the tropics is generally not correct.

Ratio 7Be/•'løPb
In the previous section we showed that 7Be/•'løPb is a sensitive tracer of vertical transport. Here we use in Figure  We showed in the previous section that the seasonal variability of 7Be/2•øPb is much larger than the spatial variability of the annual average ratio and that the model simulation of the ratio seasonality over land was fairly accurate. Figure 14 shows the 7Be/2•øPb surface concentration ratio in January and July, normalized by the annual average ratio calculated at each grid box.
This normalization removes variability associated with the radionuclide sources. Figure 14