An investigation of South Pole HOx chemistry: Comparison of model results with ISCAT observations

Unexpected high levels of OH and NO were recorded at the South Pole (SP) Atmospheric Research Observatory during the 1998–99 ISCAT field study. Model simulations suggest a major photochemical linkage between observed OH and NO. A detailed comparison of the observations with model predictions revealed good agreement for OH at NO levels between 120 and 380 pptv. However, the model tended to overestimate OH for NO levels <120 pptv, while it underestimated OH at levels >380 pptv. The reasons for these deviations appear not to involve NO directly but rather HOx radical scavenging for the low NO conditions and additional HOx sources for the high NO conditions. Because of the elevated levels of NO and highly activated HOx photochemistry, the SP was found to be a strong net source of surface ozone. It is quite likely that the strong oxidizing environment found at the South Pole extends over the entire polar plateau.


Introduction
The photochemistry of near surface air at the South Pole (SP) has received minimal attention from the modeling community in past years. This most likely reflects the fact that other than 03 and CO, little data existed to justify detailed model studies. The recent field study ISCAT (Investigation of Sulfur Chemistry in the Antarctic Troposphere) provides extensive photochemical observations so that a comprehensive modeling analysis can be carried out to investigate SP summertime photochemical processes. The ISCAT program represents an extension of the earlier SCATE program (Sulfur Chemistry in the Antarctic Troposphere Experiment) that examined sulfur chemistry at Palmer Station located on the Antarctic coast. This extension to the South Pole recognizes that the different chemical forms of sulfur found in Antarctic ice cores can serve as useful climate proxies (e.g., Legrand, 1997 and references therein.). It also highlights the continuing need to improve our quantitative understanding of the dynamical and chemical factors influencing sulfur speciation and deposition at the SP.
Among the important chemical factors is the role played by local photochemistry in altering species before their deposition. In this regard, the levels of the hydroxyl radical (OH) are most critical. The OH radical is now recognized as the single most important atmospheric species responsible for converting oceanreleased reduced sulfur (i.e., dimethyl sulfide, DMS) to its several possible oxidized forms [Hynes et al., 1986].
One of the critical species that controls OH, particularly at very low temperatures, is NO (e.g., see Figure 1). Early Antarctic coastal measurements of NO during the SCATE program led to the conclusion that SP levels would probablynot exceed 5 pptv [Berresheim et al., 1998]  All data collected during the ISCAT study involved sampling from either the second floor of the NOAA operated Atmospheric Research Observatory (ARO) building or from the roof of the ARO building. The OH and NO sampling inlets at the second floor extended out 1 -2 m from the building, at a height of-10 m above the snow surface. All sampling inlets were mounted on the side of the ARO building facing the prevailing wind, e.g., out of the "clean air sector" (0 ø-120ø).

Model Description
The photochemical box model used in this study was similar to that described previously by [Crawford et al., 1999]. The model assumes that all short lived species to be in photostationary state. The longer lived species (e.g. H202, CH3OOH, HNO3, etc) were assumed to be at steady state. Model runs were constrained by observational values of 03, NO, CO, H20, CH4, pressure, and temperature. For select sensitivity runs, the model was also constrained using measured values of OH. All model constraints were based on 10 min averaged data. The photolysis rate coefficients were based on in-situ spectrally resolved actinic flux measurements [Lefer et   However, since the HNO3 observations were very limited, the resulting "k" values for surface loss rates must be viewed as "best estimates" only. The HNO3 values themselves ranged from 120 to 750 pptv (R. Arimoto, unpublished results). The first order loss coefficients estimated from HNO3 ranged from 2.4x 10 -6 to 1.8x 10 -5 sec '•. For our standard model, we selected a first order "k" value near the higher end of this range, i.e., 1 x 10 -5 sec -• However, in the model' s current configuration, ß the maximum impact on predicted OH when using other surface loss "k" values cited was only +16%.

OH: Observations vs. Model predictions
In the text that follows, only when both OH and NO values were recorded simultaneously was an "observation" versus "model" comparison considered. In addition, comparisons were only made when temporal variations in NO over a 20 min period were < 25%. This filtering procedure resulted in a total of 316 independent 10 min. OH values, or -80% of the total OH data.

Detailed investigation of data groups L1 and L2.
An important atmospheric characteristic revealed from examining data subsets L 1 and L2 was that very high dew points were recorded for both sampling days (i.e., -25 øC for 12/17 and -29 øC for 12/27). In fact, these were the only cloudy/foggy days during ISCAT for which both model runs and observations were available. As suggested by [Mauldin et al., this issue], the erivironmental conditions on these two days points strongly toward the possibility that there were additional losses of HOx radicals due to droplet scavenging. Previous studies have reported evidence of substantial HO2 loss in clouds [Mauldin et al., 1998;Cantrell et al., 1996]. To simulate this løss in the current study, we introduced a first order "k" value for HO2 with an assigned sticking coefficient, ¾, of unity. This scavenging process was further assumed to be irreversible. Given the temperature and dew point on 12/17 and 12/27, we then tookthe supercooled water droplet size distribution as falling wRhi/t •the range of 5 -10 gm and as having an estimated number density of 5 to 15/½m 3 [Austin Hogan, unpublished results]. Using the midpoint of these ranges, the resulting first order scavenging rate was estimated at 9.0 x 10 -3 s '• [Fuchs and Sutugin, 197011 The required "k" values needed to bring the model and observations into a high level of agreement were determined to be 8.0 x 10 '3 s -• and 3.5 x 10 -3 s -• for L1 and L2, respectively.

Thus, within the range of values cited for droplet parameters, a modification of the standard model was able to bring both days'
predictions into reasonable agreement with the observations. The corresponding lifetimes for HO2 are 2.1 and 4.8 minutes. 3.1.2. Detailed Investigation of Data Group H. As discussed in the above text, the group "H" appear unique in that the observed OH is substantially higher than the standard model prediction. In this context, recent polar observations indicating snow emissions of CH20, H202, and HONO as a possible source of HOx radicals would appear to be quite relevant [McConnell et al., 1997;Sumner and Shepson, 1999;Hutterli et al., 1999;and Dibb et al., 1999]. To simulate the effect of the new sources, we have carried out sensitivity runs to estimate the level of each of these species required to remove the model underestimation. Then, the estimated levels have been compared with available data from various polar sites. In the case of CH20, additions of only 80 pptv were found to be sufficient. Observed CH20 levels On the other hand, if similar levels of these new HOx sources are applied to our group M data, the model predicted OH is increased by a factor of 1.3 to 2.5, which is significantly higher than the observed OH. By contrast, •for data groups L 1 and L2, the same addition of CH20 and H202 shifts the predicted OH up by less than 20%. HONO, however, increases predicted OH by more than a factor of•2. To reconcile these discrepancies, one has to assume that the actual levels of these additional HOx sources would to a large extent parallel NO levels, which as noted by Davis et al. [this issue] is strongly controlled by the atmosphen.'c mixing depth at SP. If true, then the general trend would be one in which the levels of all four species would be significantly modulated by shifts in the mixing depth. Thus, the OH results for data groups "M" and "L" would not be expected to be as strongly influenced by the proposed additional HOx sources as group "H". Suffice it to say, any comprehensive

SP OH and HOx Photochemical Budget
As shown in Table 1, primary production of OH from the reaction O(•D)/H2 ¸ accounts for only 6% of the total. The major source is from the recycling reaction, NO +HO2. Major OH sinks involve reaction with CO, CH4, and NO2, e.g., 51%, 20%, and 9%, respectively. The remainder is due to the reaction of OH with 03 (7%), H2 (6%) and CH20 (4%).

Consequences of Intense SP HOx Chemistry
One of the interesting consequences of the intense surface layer photochemistry at the SP is the prediction that significant net O3 should be a result. By extension, the same statement should also apply to the entire Antarctic plateau region. This finding is quite unique relative to what is typically found at a remote surface site where NO levels are usually quite low. In the latter case one normally finds net photochemical O3 destruction. Based on the SP observations, the predicated net photochemical production of 03, P(O3), is estimated to range from 1 to 6 ppbv/day (e.g., see Crawford et al. [this issue]). The major pathway for this formation involves the reaction HO2/NO (see Figure 1). SP photochemical O3 destruction ranges from 0.3 to 0.8 ppbv/day, the two most important destruction processes being OI-I/O3 and OH/NO2.
The results from the first ISCAT field study have once again demonstrated that atmospheric surprises are still be found. Quite remarkable is the finding that the near surface atmospheric layer at SP, with greatly enhanced NO levels, produces a 24 hour average oxidizing level (i.e. OH = 1.7 x 106 molec./cm 3) which rivals that of equatorial marine environments. It is also 20 times higher than the 24 hour average value estimated for Palmer Station, Antarctica [Jefferson et al., 1998].
Since for most chemical species deposition to the snow surface occurs during the summer months [Bergin et al., 1998], these results raise some interesting new questions about the degree to which some species might be modified before being deposited at the surface. In conjunction with growing evidence that extensive oxidative processes are also occurring within the snowpack (i.e., fim) [e.g., Sumner and Shepson et al., 1999], the interpretation of the concentration levels of some climate proxy species in ice cores may need to be reexamined. These results also point to the need for new research to explore the impact from surface emissions of NO and other trace gases on near surface OH levels for other snow covered regions.