CLOUDS AND WET REMOVAL AS CAUSES OF VARIABILITY IN THE TRACE-GAS COMPOSITION OF THE MARINE TROPOSPHERE

We describe a modeling study of the effects of clouds and wet removal on the chemistry of the remote marine troposphere. Using a time-dependent model with parameterized vertical transport to calculate trace-gas concentrations, we find that large variations in key species (e.g., HNO3, H2CO, and H202) result from simulations of sporadic rainfall, changes in cloud cover, and external inputs such as surface NO sources. Depending on the frequency and intensity of an event, the effects of these perturbations may persist for several days, thereby invalidating assumptions of photochemical equilibrium in the interpretation of measurements. Long-term integrations with fixed boundary conditions and regularly occurring cloud and rain episodes demonstrate a strong sensitivity of the mean concentration of longer-lived soluble gases to precipitation frequency but also confirm the validity of using properly chosen parameterizations of wet removal in steady state calculations. The marine atmosphere is represented in our model by selecting boundary conditions such as oceanic albedo and lower background NOx and hydrocarbons than observed over continents. The numerical model includes fairly complete gas-phase photochemistry, multiple scattering optics, and a simple parameterization of a marine boundary layer. Although a simple one-dimensional vertical transport is assumed, we demonstrate clearly that rainfall and cloud-cover changes contribute to species variability. Sensitivities to exchange rates of gases with the sea surface are also discussed. test removal


INTRODUCTION
Trace gas photochemistry in the marine troposphere may be highly variable in both time and space. Spatial variability, for example, between remote marine and coastal regions, may be caused by differences in background meteorology and characteristic physical and chemical processes. Temporal variability within a given environment may result from number and irregular in time and space that their interpretations depend on our understanding their allowable variations. It turns out, for example, that a short-lived perturbation may disrupt the photostationary relations among some gases for a number of days so that field measurements of these species are difficult to rationalize without a knowledge of the recent history of the air mass being sampled. changes in sea state, synoptic weather conditions, and We first describe our model of the marine atmosphere episodic alterations in chemical inputs. Realistic modeling of (section 2) and some sensitivities of background level chemmarine boundary-layer chemistry, therefore, requires time-istry to various assumptions of chemical inputs (i.e., bounddependent chemical transport, rates of photolysis, and heterogeneous removal of trace gases. The latter is very complex and includes gas-particle and air-sea transfer, chemical transformations within aerosols, clouds, rainwater, and on the sea surface.
We report here the results of a modeling study that considers one aspect of the variability due to heterogeneous processes, namely, the influence of sporadic clouds and precipitation removal on trace gas distributions in the marine troposphere. Using a one-dimensional time-dependent transport-kinetics model to simulate these processes, we find that in a given air mass the concentrations of key species, particularly soluble gases such as HNO3, CH20, and the peroxides, may change by a factor of 5-10 in a few hours as a consequence of washout below a precipitating cloud. Further, depending on the mean cloud and precipitation frequency of a given environment, the spatial variability of these same compounds may also approach an order of magnitude. Even without wet removal an extended period of cloudiness may cause substantial changes in trace gas distributions.
These results are not simply of theoretical interest because ary conditions) (section 3). Sensitivities arise because onedimensional transport restricts time-dependent calculations to a fairly uniform air mass and time-dependent simulations are always started from an initial condition with a fixed ozone profile and corresponding diurnal behavior for the other gases. In section 4 we consider two basic types of synoptic events, focusing on their effects on odd nitrogen and soluble gases. In section 5, assuming that certain synoptic episodes are regular events, we present the results of long-term integrations to look at effects on ozone and to test steady state assumptions of wet and dry removal of soluble gases.

MODEL
Chemical distributions of trace gases are determined as a function of altitude from a system of one-dimensional transport-kinetics equations where the time rate of change of species j is the sum of its chemical reaction rates and flux divergence, represented by eddy diffusion:

K(z) N(z) + Pj(z, t) -Lj(z, t) = Ocj(z, t)
Oz Oz ot (1) Profiles for temperature, 02, N2, and H20 vapor taken from U.S. Standard Atmosphere (1976). *Mixing ratios for CO, CH4, and H2 are fixed at 15 and 0 km and are calculated from a steady state code to initialize time-dependent calculations. Values shown correspond to model 4 (see Table 3). For all other model boundary conditions, mixing ratios for CO vary up to 15% of values given here but do not change for H2 and CH4.
We simulate the marine atmosphere by assuming fixed profiles for N2, 02, temperature, CH4, CO, H2, and H20 (Table 1) and solve the system of continuity equations (1) for the volume mixing ratios of 17 gaseous species: 03, O(3p), NO, NO2, NO3, N205, HNO3, HNO4, H, OH, HO2, H202 and the hydrocarbons resulting from methane oxidation CH3, CH30, CH302, CH3OOH, and CH20. For convenience, reference is made to collective terms: odd oxygen, Ox = 03 + O(3p); odd nitrogen, NOy = NO + NO2 + HNO3 q-2N205 + HNO4 + NO3; odd hydrogen, HOx = H + OH + HO2; however, mixing ratios are always calculated for all the individuals, not the 'group' as a whole. We point out that NOy as we define it does not include peroxyacetyl nitrate (PAN) which may be a major form of odd nitrogen in the mid-to-upper troposphere [Singh and Hanst, 1981]; our NOy also excludes a contribution of approximately 200 ppt due to HCN from ground-level sources [Coffey et al., 1981;Cicerone, 1982]. Over 50 gas-phase radical reactions and photolyses describe the kinetics of free radicals and more stable molecules; these appear in Table 2, where most of the rates are taken from the recent NASA Panel Evaluation [NASA, ary layer and a surface boundary layer. In the standard parameterization of eddy diffusion shown in Figure 1, it is assumed that the planetary boundary layer (PBL) (characteristic diffusion coefficients, K = 105-106 cm 2 s -l) extends to the ground. A separate surface boundary layer (SBL), a region of reduced mixing between the ground and about 50 m altitude, is simulated with a value of K(z), 2-3 orders of magnitude lower than in the planetary boundary layer and which increases with altitude. [see, e.g., Businger, 1973]. The shape of K(z) was chosen to give a maximum value that corresponds to a mixing period of roughly 20 min in the several hundred meters above the surface boundary layer.
A single diurnal cycle consists of 15 time steps at night and 22 daytime points (solar zenith angles), with the majority clustered around sunrise and sunset, minimum At = 2 min.
One advantage to solving the full continuity equations (2) for individual species, rather than only for chemical 'families' like NOy, HOx, is that short-lived radicals (e.g., O, OH, CH20, NO) are always mathematically uncoupled from their longer-lived precursors. Similarly, highly soluble species like HNO3 are mathematically uncoupled from less soluble species such as NO. Thus, we do not depend on assumptions of chemical equilibrium.
For sensitivity studies (e.g., to examine the effects of varying boundary conditions and reaction rates) a steady state version of the model is used to solve equations (2) for the case dx/dt = 0. In the steady state model, diurnally averaged species concentrations are calculated according to the procedure of Turco and Whirten [1978]. Briefly, one multiplies each chemical rate constant by a factor that contains information on the diurnal behavior of the reactants. To compute these factors, the time-dependent code must be integrated until each species displays a 24-hour cycle and ozone (the most slowly varying of the chemical constituents) has converged to a fixed profile. In this manner, the steady state model calculates 24-hour averages for all species. The steady state and diurnally averaged timedependent codes give results within 5% for all species at all altitudes.

fixed concentration, (2) specified flux (•, in molec cm -2 s -l) or surface removal velocity (Ure m in cm s-l), and (3) photochemical equilibrium that can be assumed when the concen-
The standard parameterization for each model (Table 3) has been chosen to reproduce concentrations of ozone and

NOy consistent with the limited data available for the remote low-to mid-latitude marine boundary layer. These include measurements made as part of GAMETAG (Routhier et al. [1980] for 03; Huebert and Lazrus [1980] for HNO3) and
tration of a short-lived radical is essentially independent of other recent determinations of odd nitrogen that suggest as a transport. The overall chemistry is very sensitive to the forcing from boundary conditions for the longer-lived gases. Ozone and NOx (=NO + NO2) are critical in this respect, but it is difficult to specify values with certainty because only a few isolated measurements of these species have been made in the marine troposphere, especially near the sea surface. Table 3  Kink between 2 and 3 km in cloud case results from assumption of a single dense cloud layer. Column-integrated net 03 production, 5-15 km: 2.9 x 101ø cm -2 S -l, no cloud; 4.3 x 10 lø cm -2 s -1, cloud. (8) Low values of NOy and 03 are compatible only with fairly rapid NOy deposition. For example, if a uniform removal velocity is specified for all NOy species (this is equivalent to a grouped or family approach), total NOy on the order of 50-60 pptv requires a transfer rate, v = 0.20 cm s-i. In that case the NOy individuals, NO, NO2, NO3, etc., are in approximate photochemical equilibrium relative to one another, with calculated HNO3 making up 80% of the total (Figure 6). McFarland et al., 1979;Huebert, 1980] implies that this fraction is too high and HNO3/NOx is probably closer to 1-3. By varying the deposition velocity for each species in turn it is possible to obtain a slightly greater porportion of NO2, but for a given total NOy there is only a small range over which NOx/NOy can be changed by this approach.

An examination of recent marine data [Helds and Warneck
Thus, it is likely that near the surface odd nitrogen chemistry is dominated not by surface removal but by other processes in the boundary layer or even the free troposphere, for example, heterogeneous loss of HNO3 [Fishman and Crutzen, 1977], nonstratospheric sources of NOx such as lightning [Chameides et al., 1977;Noxon, 1976], transport from continental areas [Chameides, 1978], or a local source [Zafiriou and McFarland, 1981] (Table 3).  Figure 8 where the solid curve represents the mixing ratio calculated assuming efficient mixing to the ocean surface and the dashed one has been computed from reduced nearsurface transport. (Although they are not illustrated here, a similar pattern applies to the other soluble gases, HeCO, H:O:, and CH3OOH.) We would have to specify a very high surface removal velocity along with surface resistance (approximately 3 times greater than for the no-SBL case) to bring the dashed curve for HNO3 into coincidence with the reference distribution in Figure 8. Providing that real-time determinations of a soluble gas are possible, the 'bottleneck effect' should be detectable when field measurements are made under conditions of reduced transport. If enhanced levels of these species are not observed, there is probably very effective destruction at the sea surface or heterogeneous removal by other processes in the boundary layer (e.g., removal to sea-salt particles or fog drops [Heikes and Thompson, 1981]). This simple example of chemical sensitivity to transport argues strongly for an accurate knowledge of dynamics when interpreting measured gas distributions. Note, for example, that the ratio [NOx]/[HNO3] varies strongly with altitude below 100 m in Figure 8. It may be that in a regime characterized by two or three layers of varying mean vertical diffusivity, there is a stratification of background chemistry, although horizontal transport could counteract these effects that have been simulated here only in one dimension.

The second effect of a resistant surface boundary layer is to introduce a similar bottleneck into the surface deposition of the longer-lived soluble gases, causing them to accumulate in the planetary boundary layer. This phenomenon is depicted for HNO3 in
In describing steady state results, we have tried to illustrate a few sensitivities of background level (low NOy, low 03) marine tropospheric chemistry which might be observed on a short time scale as chemical inputs and transport vary in a given air mass. These results might characterize a particular marine environment that has been stable over a long period of time. have also performed a model run with continuous NO upflux (7.5 x 108 cm -2 s -2 day and night). The only differences in NOy mixing ratios between that run and model 1 are for NO itself (which instead of disappearing completely at night falls to a few percent of its daytime maximum) and for NO:, which increases 15-20% in response to NO forcing at night, although only 5% on a diurnally averaged basis. The constant NO input also eliminates the small diurnal variation in total NOy (cf. Figure 9a).

TIME-DEPENDENT CALCULATIONS AND SHORT-TERM
We note that although our calculated NOx and HNO3 at surface are consistent with the data available, measurements of NO3 in unpolluted air [Noxon et al., 1980;Platt et al., 1981] Table 3), which suppresses nearsurface NO3 relative to NO2. This is equivalent to substantial heterogeneous removal, a likely cause for low NO3 [Thompson and He&es, 1982]. Thus, our calculated NO3 may be fairly accurate for the background marine troposphere. Figures 10a and 10b Figure 11; these species increase slightly over the course of a day as HOx accumulates. In models 1 and 2, NO forcing consumes more HO2 via reaction  1 and 4-A) were selected to illustrate this sensitivity.

and 4-A are given in
For simplicity we emphasize qualitative aspects of synoptic effects and restrict simulations to two prototypical events. In the first type a 3-day period of cloudiness is represented by including the simple cloud described above in the calculation of photolysis rates; this gives a measure of chemical sensitivity to changes in the radiation field. A second type of episode is a precipitation event in which the cloud effect on radiation is combined with below-cloud washout of soluble gases. Figure 12. Reduced photolysis below cloud decreases the concentration of the primary HOx radicals, hence the rate of formation of HNO3, H202, and the products of methane oxidation. The magnitude of the change is proportional to the radiation blockage (i.e., to cloud albedo, 0.8 in these simulations) and also depends on the lower boundary conditions. For example, when a NO upflux is assumed (model 1), at the end of 3 days the HNO3 mixing ratio has declined 8% below its initial value, but with less conversion of NO2 to HNO3, NO• has increased 20%. One might suppose, however, that the NO upflux, which is derived from a photochemically driven source in the sea, is reduced during cloud cover in proportion to the decrease in radiation intensity at the surface (i.e., 80%). A 3-day cloud, combined with this altered boundary condition (model Cloud-2 in Figure 12), suppresses NO• and HNO3 until at the end of 3 days the latter has decreased to only 75% of its no-cloud value. For a system in which the initial distribution of N Oy species is already near photochemical equilibrium, cloud-induced effects are negligible for NO• and HNO3 (model 4-A, Figure 15a) but cause a 10-15% decrease in H202 and the hydrocarbons (Figure 13). In all cases illustrated here, the removal of the cloud at the end of 3 days along with the restoration of the original boundary conditions, gives rapid return toward initial values. This is evidence that although radiation effects may be significant, The assumption of irreversibility in precipitation scavenging is valid for species that react rapidly and essentially completely in solution (e.g., a strong acid):

HNO3,aq ' • H(aq) + q-NO3,aq ,-Keq > 1
Nitric acid is the only one of our scavenged gases for which this assumption is clearly justified. However, it seems reasonable to suppose that irreversible washout also applies to other species that react with water or another solute fast enough (typically <102 -103 s) that the falling raindrop does not become saturated. For example, formaldehyde hydrates (H2CO lifetime against hydration 0.1 s [Schecker and Schulz, 1969]) and the peroxides can be consumed as they oxidize HSO3- [Penkerr et al., 1979;Heikes et al., 1982]. For species for which saturation effects in scavenging are significant, k•,r decreases with distance from the cloud, and the corresponding vertical profile sequence shows at all times a minimum mixing ratio just below cloud level where precipitation washout is most effective, 1-2 km in our simulations. These results assume that K(z) does not change (i.e., transport is constant over the course of the event). If the precipitation event begins from an initial condition based on the transport scheme of a resistant surface boundary layer (model 1-SBL; cf. Figure 8) the moving picture of a 4-hour episode for HNO3 is shown in Figure 14b with the solid line representing the initial profile. It was pointed out (section 3) that only a very high rate of surface deposition or heterogeneous removal in the planetary boundary layer could bring the no-cloud, no-rain curve calculated with a surface boundary layer into coincidence with one computed assuming rapid mixing to the ground (dash-dotted line in Figure 14b). Precipitation washout can be an efficient mechanism for accomplishing this; with a wet removal coefficient kwr = 2 x 10 -4 s -1, it takes less than an hour to reduce HNO3 in the planetary boundary layer to the level of the no-SBL distribution.
The time dependence of soluble-gas mixing ratios at the surface during and after a precipitation event is illustrated in Figures 15a and 15b. The no-cloud, no-rain reference diurnal cycle is the solid line near the top, and several variations of a simple cloud-and-rain perturbation are simulated (Table 5). Model Rain-1 consists of a single cloud-and-rain pulse lasting 4 hours, beginning at sunrise on model day 0. In Figure 15b this corresponds to a time in which HNO3 is still increasing in its normal diurnal pattern under the influence of an NO upflux. If no further perturbation occurs and the same lower boundary condition is assumed, a damped diurnal pattern sets in as the original photochemistry and transport are reestablished. Within a few hours, HNO3 is restored to 25% of its normal no-rain value; and at the end of a week, 90% of the original HNO3 level is achieved. When there is no NO input (Figure 15a), recovery of HNO3 from a 4-hour event is much slower. In both cases the NOx diurnal patterns are virtually constant over the course of 7 days. In the simulation Rain-2 (Figure 15b), if the cloud cover of the storm produces a proportional reduction in the surface upflux of NO (80% loss), a 4-hour rain pulse reduces HNO3 --•25% more than in Rain-l; after a day or two of integration, however, the concentration from models Rain-1 and -2 are virtually identical. A similar phenomenon is observed if the initial perturbation lasts 8 hours instead of 4; cf. Rain-1 and Rain-4 in Figure 15b. This demonstrates the dominant influence of the NOy source (cf. Figure 12). In model Rain-3, a second 4-hour cloud-and-rain shower takes place on model day 3, greatly delaying restoration of soluble gas concentrations to their prerain levels (Figures 15a and 15b).
The sensitivity of surface chemistry to the various episodes that have been simulated here is summarized in Table   5, where for NOx and the soluble gases HNO3 and H2CO (representing, maximum and minimum wet removal efficiency, respectively) we have computed the ratio of the mixing ratios on model days 3 and 7 to the initial value, day 0 (i.e., the fractional recovery from the start of the event). Models Cloud-1 and Cloud-2 refer to cloud-only perturbations that have maximum effect after 3 days; their effect 1 week after the start of the episode is relatively small for HNO3 and  Fig. 15b. Effect of precipitation removal on surface HNO3 mixing ratio (model 1 initial condition) rain models 1-4 are described in text and Table 5. with only one third the wet removal rate of HNO3, effects at the 3-day point are significant, but at the end of a week recovery is nearly complete. The critical dependence of these results on boundary conditions cannot be overemphasized. Except in specified NOx in two model runs, a fixed set of boundary conditions has been maintained for a week of integration in all of these simulations. Hence, our results apply only to extremely stable meteorological conditions; and our numbers, especially for the recovery time of soluble gases, must be regarded as rough limits for what might be found in the atmosphere. Nonetheless, we believe we have made a good estimate of the magnitude of short-term variability which typical synoptic episodes might impose in the boundary layer while identifying other key sensitivities of background-level chemistry. These model results have two practical implications for measurements of trace gases in the marine atmosphere.
First, if precipitation washout of a soluble gas reduces its concentration by a factor of 5-10 in just a few hours, the change should be detectable if the time scale of measurement is kept small and if offsetting effects of advection and convection are insignificant. Such washout has been observed in polluted air for H2CO [Platt et al., 1979], and recently for HNO3 removal by snow at a clean continental site (B. J. Huebert, personal communication, 1981) but confirming data in the marine troposphere are lacking. Measurements of gas-phase HNO3 in the Equatorial Pacific surface layer showed highest HNO3 levels during a storm [Huebert, 1980]; it seems most probable that storm-induced convection introduced air enriched in NOy into the area being sampled.
Second, the assumptions of photochemical equilibrium, which are sometimes used to interpret field data, are inadequate in the marine boundary layer (and probably over continents also) when applied to soluble or surface-active gases. For example, in the absence of heterogeneous reac-tion if transport is slow relative to chemistry the NO2/HNO3 ratio is given by the expression where J4 is the photolysis rate of HNO3, k36 is its bimolecular formation rate from reaction of NO2 and OH, and k37 is the rate of its loss to reaction with OH. If surface HNO3 changes 10 fold over the course of a cloud-and-rain episode (Figures 15a and 15b) Figures 15a and   15b). Thus, if one were to measure HNO3 and NO2 simultaneously at some time after the event, the ratio NO2/HNO3 could not be interpreted without knowing something of the recent history of the air mass being sampled. Finally, the different surface reactivities of species such as NO2 and HNO3 can produce strong altitude gradients in the ratio of their concentrations in the surface boundary layer (see Figure 8).

TIME-DEPENDENT AND STEADY STATE SIMULATIONS:
LONG-TERM VARIABILITY Two long-term integrations have been performed assuming regular patterns of cloud and precipitation to determine effects on ozone and other longer-lived gases and to verify the validity of conventional parameterizations of precipitation removal.
We simulate the effects of prolonged cloud cover by continuing the integration begun with a 3-day cloud (model 4 initial condition) until a new steady state ozone profile is obtained. Below-cloud mixing ratios of the longer-lived soluble gases decline up to 20% with loss of radical formation depending on the compound and lower boundary condition (Table 6). Ozone increases --•20% throughout the troposphere; much of the increase is due to greater photochemical production in the upper troposphere (Figure 3) where radiation reflected from the cloud enhances net ozone production over the no-cloud starting point [Thompson, 1980]. Because of the long time required to achieve equilibrium, the ozone variation cannot be confirmed easily by observation, but the calculation may point to important differences in background chemistry between predominantly cloudy and cloudless environments.
A second long-term integration conducted from a nocloud, no-rain initial condition (model 1) simulates the effect of periodic washout of soluble gases assuming that a 4-hour rain shower (model rain-l) occurs once a week beginning at sunrise. Slowly varying transients induced by periodic perturbation eventually produce a new converged ozone profile and a weekly cycle for each gas. Seven-day cycles are illustrated for HNO3 and H2CO in Figure 16; weekly mean surface mixing ratios, calculated for these species and the peroxides, appear in Table 7. Mixing ratios are also calculated from the steady state model by using the inadequate but standard practice of parameterizing wet removal by adding a constant removal rate •:ro to diurnally averaged photolysis and radical reaction terms for the soluble gases (e.g., for HNO3)

SUMMARY AND CONCLUSION
A one-dimensional time-dependent chemical-transport model of the unpolluted marine atmosphere has been described along with some steady state results that illustrate principal sensitivities of boundary-layer chemistry to variations in surface NOy, air-sea exchange rates, and nearsurface transport. Simulations of sporadic clouds and rainfall superimposed on the background chemistry of a fixed air mass create large short-term changes (up to an order of magnitude) in the concentrations of soluble gases. As a consequence, the photochemical equilibrium of these compounds may be upset for a period of some days, an effect that cannot be neglected in trying to interpret measurements of these species. By choosing a few representative cases of regular cloud and precipitation removal we have tried to illustrate the sensitivity of lower troposphere chemistry to differences in average meteorology. Depending on the mean cloud and precipitation patterns of a given area, the spatial variability of soluble gases over a region of uniform background chemistry may also approach an order of magnitude.
In conclusion, we emphasize that although we have demonstrated a strong sensitivity of boundary-layer chemistry to typical synoptic episodes, the exact magnitude of such variations depends critically on dynamics and chemical inputs that have been simulated here in only one dimension neglecting cloud transport. Also, in focusing on precipitation washout as a primary mechanism for perturbing chemical paper. We appreciate the comments of S. E. Schwartz and two anonymous reviewers.