A reassessment of HO x South Pole chemistry based on observations recorded during ISCAT 2000

Reported here are modeling results based on ISCAT (Investigation of Sulfur Chemistry of Antarctic Troposphere) 2000 observations recorded at the South Pole (SP) during the Austral Summer of 2000. The observations included a comprehensive set of photochemical parameters, e.g., NO, O 3 , and CO. It is worthy to note that not only were OH and HO 2 observed, but also HO x precursor species CH 2 O, H 2 O 2 , and HONO were measured. Previous studies have suggested that HONO is the major source of OH/HO x in the Arctic; however, observed HONO levels at SP induced dramatic model overprediction of both HO x and NO x when used to constrain the model calculations. In contrast, model predictions constrained by observed values of CH 2 O and H 2 O 2 are consistent with the observations of OH and HO 2 (i.e., within 20%) for more than half of the data. Signiﬁcant model overpredictions of OH, however, were seen at the NO levels lower than 50 pptv or higher than 150 pptv. An analysis of HO x budget at the median NO level suggests that snow emissions of H 2 O 2 and CH 2 O are the single most important primary source of SP HO x , contributing 46% to the total source. Major sinks for HO x are found to be dry deposition of HO 2 NO 2 and HNO 3 as well as their reactions with OH. Although ISCAT 2000 led to a major progress in our understanding of SP HO x chemistry, critical aspects of this chemistry are still in need of further investigation.


Introduction
It has long been recognized that free radicals are responsible for the tropospheric oxidation of most reduced trace gases emitted from natural as well as anthropogenic sources (e.g., Levy, 1974;Logan et al., 1981;Chameides and Davis, 1982;Thompson, 1992). Among the more important of these is the hydroxyl radical (OH). It largely controls the oxidizing power of the atmosphere. Yet another is the hydroperoxyl radical (HO 2 ), a product of many OH reactions. In the presence of nitric oxide (NO), rapid cycling between OH and HO 2 can enhance atmospheric OH through secondary production strengthening its dominance as the major tropospheric oxidizing agent. Because of this tight chemical linkage, the sum of HO 2 and OH is frequently abbreviated as HO x . Studies focused on tropospheric HO x sources and sinks are numerous and define one of the most important aspects of atmospheric chemistry (e.g., Crawford et al., 1999;Chen et al., 2001a, b;Jaegle et al., 2000Jaegle et al., ,2001Olson et al., 2001;Tan et al., 1998Tan et al., , 2001Wang et al., 2001).
Our present understanding of HO x chemistry indicates that at low altitudes production of HO x is often proportional to UV solar radiation and H 2 O levels. This naturally leads to an expected strong latitudinal gradient in OH. For example, both model results and field observations reveal that the tropics have some of the highest OH values to be found (e.g., Logan et al., 1981;Chamedies and Tan, 1981;Mauldin et al., 1999). For this reason, it was with great surprise that the first South Pole (SP) observations revealed OH levels whose summertime 24 h average was within 10% of the corresponding value in the tropics (Mauldin et al., 2001). Concomitant with these elevated OH levels were extraordinarily high mixing ratios (median, 225 pptv) of NO (Davis et al., 2001). These authors presented strong evidence suggesting the source of this SP NO being emissions from the snowpack, a by-product from the photolysis of nitrate. A modeling analysis of these ISCAT 1998 (Investigation of Sulfur Chemistry in Antarctic Troposphere) data by Chen et al. (2001b) showed that OH levels were a strong non-linear function of NO. The dominant source of OH was the recycling of HO x through the reaction sequence HO 2 +NO-OH+NO 2 , followed by OH+CO(O 2 )-HO 2 +CO 2 (e.g., see Table 1 [R4] and [R3]). It was further shown that the largest primary source of HO 2 was photolysis of CH 2 O [R13], the latter being a product from the OH initiated oxidation of CH 4 [R10]. Due to the high NO levels the amplification of HO x through CH 4 oxidation was nearly a factor of 1.5 (i.e., HO 2 produced per OH consumed). The oxidation of CH 4 , therefore, serves as a net source of HO x , rather than a net sink as is typical of most remote boundary layer (BL) settings.
While the observed trends in NO, H 2 O, O 3 , and UV irradiance appeared to explain a large fraction of the ISCAT 1998 OH observations, a more detailed comparison of model predictions against observations exposed some shortcomings. Among these was the underprediction of OH at certain times. This has suggested the possibility of additional HO x sources. Based on several Arctic studies, the additional sources were speculated to be the result of snowpack emissions of H 2 O 2 , CH 2 O, and/or HONO (e.g., Dibb et al., 2002;Hutterli et al., 1999Hutterli et al., , 2001McConnell et al., 1997;Sumner and Shepson, 1999;Zhou et al., 2001). Sensitivity tests by Chen et al. (2001b) showed that the photolysis of only very modest emissions of any of the above species could have provided the necessary primary source of HO x . However, it was not possible to quantify these potential sources since no measurements were made.
From Chen et al.'s (2001b) analysis, it was also shown that the major HO x loss pathways at SP are reactions of OH and HO 2 with NO 2 to form HNO 3 and HO 2 NO 2 . The final step in the loss process was speculated to involve mainly dry deposition of these acidic species (see Table 1

Model descriptions
The model used in this study is the same as that employed in the ISCAT 1998 analysis. It is a time dependent box model containing explicit HO x -NO x -CH 4 chemistry (71 reactions) and parameterized NMHC chemistry (184 reactions). The latter chemistry has been significantly modified from the CAL scheme reported by Lurmann et al. (1986) (see e.g., Crawford et al., 1999). Typical model inputs consist of NO, CO, O 3 , H 2 O, ambient temperature, and pressure. However, the model can also accommodate observational constraints based on measurements of OH, HO 2 , HNO 3 , HO 2 NO 2 , H 2 O 2 , CH 2 O, and HONO. The photolysis coefficients used in the analysis are those derived from in situ actinic flux measurements (see Davis et al., 2004a). Since no diurnal UV flux variations occur at SP during the Austral summer (there are variations, however, due to shifts in the overhead O 3 column density), photochemical steady state was assumed for all model calculated species. As the lifetimes of HO x species are quite short, i.e., B3 s for OH and B30-80 s for HO 2 , this assumption is quite reasonable for both species. The first order rate for heterogeneous removal of soluble species was taken to be 9 Â 10 À5 s À1 , the best deposition loss rate estimate for HNO 3 and HO 2 NO 2 by Slusher et al. (2002) based on ISCAT 2000 dataset. This value is about one order of magnitude higher than that used in the 1998 study, and is based on the 2000 measurements of HNO 3 . If the current deposition loss rate were applied to the ISCAT 98 analysis, the model OH would decrease, on average, by nearly a factor of 1.5 and the median P(O 3 ) would be reduced by B0.8 ppbv day À1 . Model uncertainty based solely on rate coefficient uncertainties has been evaluated based on Monte Carlo sensitivity calculations. The model HO x uncertainty based on these calculations was estimated at 40%.

Results and discussion
A summary of the observational data recorded during ISCAT 2000, together with a description of the instrumental technique (e.g., sensitivity and calibration), are presented in Davis and Eisele, 2004. For further details concerning a specific measurement, the reader is encouraged to examine individual contributions to this special issue. For completeness, however, the typical total uncertainty (random plus systematic errors) assigned to observations of OH, HO 2 , HNO 3 , and HO 2 NO 2 were in the 20-60% range depending on the S/N (signal to noise) ratio of a specific measurement.
The interpretive analysis of ISCAT 2000 HO x observations was based on 4604 individual model runs. These runs were mostly derived from 10-min data averages of the observed parameters NO, O 3 , H 2 O, and several j values. In addition, values for CO and nonmethane hydrocarbons (NMHC) were interpolated from grab sample observations having a typical temporal resolution of 2 to 4 samples day À1 . (The authors note that the day-to-day variations in CO were typically less than 5% and NMHC levels were sufficiently low as to have a negligible effect on the HO x results.) Among the 4604 model runs completed, 2455 of these had OH measurements and 199 had HO 2 data. However, coincident OH and HO 2 data were not available because the same instrument was used to measure both species such that these measurements could not be made simultaneously. In addition to the HO x measurements, there were 754 and 1185 runs having measurements of CH 2 O and H 2 O 2 , respectively. Both CH 2 O and H 2 O 2 observations were recorded at 1 m above the snow surface (Hutterli et al., 2004). The measurement of HONO was made near the surface and at 10 m above the surface (Dibb et al., 2004). However, only very limited observations were recorded at 10 m. It was this group of data at 10 m that was used to constrain the model for purposes of evaluating the impact of HONO on SP HO x sources. Other than H 2 O 2 , CH 2 O, and HONO, all data were collected at 10 m above the snow surface. Since the vertical mixing is typically weak in SP (Oncley et al., 2004), the H 2 O 2 and CH 2 O levels may be lower at 10 m. Estimates of the eddy-diffusion calculation indicate that for neutral conditions this minor difference in sampling heights could lead to 7% and 5% drops for H 2 O 2 and CH 2 O, respectively (also see Hutterli et al., 2004). One should expect larger drops for stable conditions, which is probably more prevalent in SP. This will be further examined in terms of their impact on HO x later in the text. As related to the HO 2 observations, only the sum of the concentrations of HO 2 , OH, CH 3 O 2 , and RO 2 (Mauldin et al., 2004) could be measured as the instrument was configured at SP. The HO 2 fraction of the total radical sum was estimated from model runs. The HO 2 fraction typically ranged from 69% to 78%, with a median of 75%.

Comparison of model results with observations
As stated earlier, a major goal of this study was to quantitatively re-evaluate sources and sinks of SP mixed layer HO x under summertime conditions. In particular, this required our examining the magnitude of the HO x source resulting from snow emissions of H 2 O 2 , CH 2 O, and HONO. During ISCAT 2000, snow emissions significantly enhanced the measured levels of all these HO x precursors (Dibb et al., 2004;Hutterli et al., 2004). Our standard model (gas phase chemistry only) can only predict less than 4%, 39%, and 5% of observed levels of H 2 O 2 , CH 2 O, and HONO, respectively. Thus, assessing the effect of snow emission on SP HO x levels requires model calculations constrained by the observed values of these precursor species. The effect of snow emissions then can be estimated by comparing the constrained model results with those from the standard model. Finally, the results from both models are compared against the observations. Fig. 1a shows observed OH values and model results from the standard model. Fig. 1a shows that both modeled and observed OH are strongly modulated by NO with peak values occurring at B100 pptv of NO. It is also apparent that model values are generally lower, except for values of NO o40 pptv. Overall, the median ratio of model to observation (M/O) is 0.68. Following the same approach for HO 2 , Fig. 1b shows that observed HO 2 values are also higher than those predicted by the model. (The smaller range of NO values plotted against HO 2 versus OH primarily reflects the limited number of ISCAT 2000 HO 2 observations.) Again, the median M/ O ratio for HO 2 is about 0.65. These differences between model and observations appear to be modest in comparison with model and observation uncertainties of 40%, which suggests that the median model bias could be removed by significant shifts in some kinetic rate coefficients and/or HO x measurement calibration factors. Alternatively, the model underprediction of HO x could also indicate that the model either underestimates HO x sources or overestimates its sinks.
To test the hypothesis that HO x sources might be underestimated, the ISCAT 2000 observations of CH 2 O, H 2 O 2 , and HONO were used to constrain the model. All three species have previously been identified at Arctic sites. In each case, it has been stated that the major sources of these species were emissions from the snowpack (e.g., Dibb et al., 2002;Hutterli et al., 1999Hutterli et al., , 2001McConnell et al., 1997;Sumner and Shepson, 1999;Zhou et al., 2001). Fig. 2     constrain the model. It can be seen in the figures that constraining all three precursors leads to significant model overestimation of both OH and HO 2 . The model values are factors of 2 to 3 and factors of 3 to 5 higher than observations of HO 2 and OH, respectively. At the median NO level of 89 pptv, the model predicted OH concentration approaches to 1.1 Â 10 7 molec cm À3 . Although direct measurements of OH have not been conducted at any Arctic sites, model estimates of OH based on HONO measurements are available for Summit, Greenland and Alert, Canada. It should first be noted, however, that the average SP HONO level is about a factor of 3-4 higher than that of Summit and a factor of 6 higher than the value measured at 5 m above surface in Alert. For springtime conditions at Alert, Zhou et al. (2001) suggested that production from HONO would increase peak OH values (noon) to 1.1 Â 10 6 molec cm À3 . Based on observations at Summit, Yang et al. (2002) predicted noontime OH values approaching to 9 Â 10 6 molec cm À3 in one case; however, the derived diel average was only B4 Â 10 6 molec cm À3 . Although there are significant differences in the levels of other OH precursors between the SP and Arctic studies; the major factor driving the differences in model OH levels is most likely to be the in observed HONO levels.
In another assessment of the HONO observations, we examined the impact of HONO on the levels of NO x , the rational being that HONO photolysis [R17] also serves as a source for NO. In fact, for periods of concurrent observation, NO values (at 10 m) and HONO values observed near the surface (85 cm) are positively correlated. This relationship can be well approximated by a second order polynomial with an R 2 value of 0.83. However, it is difficult to reconcile the observations of both of these species. For example, under the conditions of ISCAT 2000, the typical photochemical lifetime of HONO was only 7-10 min while that for NO x was 7-25 h. Taking the average lifetime for each species and the median value of HONO sampled at 10 m (i.e., B26 pptv), the median NO x level is predicted to be B10 times higher than what was observed. This raises the question of whether the measurement of HONO might suffer from some systematic error (e.g., chemical interference), although one can not rule out the possibility that the model mechanism might miss major HO x and NO x loss processes unique to the polar troposphere. While the mist chamber technique has been used in numerous studies involving various environmental conditions, there still exists the possibility of interference from any species that readily dissolves in aqueous solution to make nitrite. At this point, there are major difficulties in interpreting the HONO observations. Thus, we have focused our attention on the impact from snow emissions of CH 2 O and H 2 O 2 in the text that follows. alone. Similar to other soluble species, the model assumes a dry deposition rate of 9 Â 10 À5 s À1 . The flux measurements by Hutterli et al. (2004) indicate that the snow, on average, is a net source of atmospheric H 2 O 2 and CH 2 O. But, it does not rule out dry deposition as a part of the air-snow exchange processes. The zero deposition case described by Hutterli et al. (2004) can be considered as a limiting case in which the emissions balance deposition. On average, the gas phase chemistry can only account for up to 3% and 30% of the observations of H 2 O 2 and CH 2 O, respectively. Even if the dry deposition rate is one order of magnitude smaller, it still leaves more than 70% of H 2 O 2 and 30% of CH 2 O observations unexplained. These results are consistent with those from a more detailed analysis by Hutterli et al. (2004). Together, the discrepancy between observations and model predictions based on gas phase chemistry alone argues that substantial amount of H 2 O 2 and CH 2 O are sustained by snowpack emissions. This is supported by CH 2 O air-snow transfer model predictions (Hutterli et al., 2002) and by direct CH 2 O and H 2 O 2 flux measurements during ISCAT 2000. A comparison of those fluxes with the photochemical modeling can be found in Hutterli et al. (2004). Similar results were also reported by Hutterli et al. (1999Hutterli et al. ( , 2001 for analyses of CH 2 O and H 2 O 2 observations recorded at Summit, Greenland and by Sumner et al. (2002) for a CH 2 O study in Alert, Nunavut, Canada.
To better understand the role of snow emissions, it is important to quantify how each of these emitted species influences HO x levels. As noted earlier, sensitivity calculations were based on median conditions for low, mid, and high NO levels. The results of the sensitivity calculations are summarized in Fig. 4. Shown here are the observed medians for OH and HO 2 , as well as the model predicted concentrations for these species based on four different constraint levels. Level 1 constraint represents the standard model (i.e., not constrained by either H 2 O 2 or CH 2 O); level 2 is constrained by H 2 O 2 only; level 3 is constrained by CH 2 O only; and level 4 shows the combined effects of being constrained by both H 2 O 2 and CH 2 O. From Fig. 4a, model predicted increases in OH from H 2 O 2 only range from only 0.3 to 1.0 Â 10 6 molec cm À3 (32-58%), while the corresponding HO 2 increase shown in Fig. 4b varies from 0.5 to 1.5 Â 10 7 molec cm À3 (28-60%). In the case of constraining CH 2 O only, from the lowest NO case to highest case, the model predict OH increase varies from 0.2 to 1.7 Â 10 6 molec cm À3 (21-106%), while the corresponding HO 2 enhancement ranges from 1.2 to  1.9 Â 10 7 molec cm À3 (31-134%). Considering that the observed levels of H 2 O 2 are much higher than that of CH 2 O, one can conclude SP HO x is much more responsive to CH 2 O than to H 2 O 2 on a per-molecule basis. For example, the average OH sensitivities to H 2 O 2 and CH 2 O are 3.5 Â 10 3 and 8.3 Â 10 3 molec cm À3 pptv À1 , respectively. Even larger contrast can be seen in the case of HO 2 , i.e., 0.5 vs. 1.7 Â 10 5 molec cm À3 pptv À1 . A major reason for this is the difference in the photolysis rate for the two compounds. On average, CH 2 O photolyzes nearly 5 times faster than for H 2 O 2 . Although not shown, the sensitivity calculations also reveal that the OH sensitivity is a strong function of NO, increasing by a factor of 4 from the lowest NO level to the highest as the partitioning of HO x is shifted more favorably toward OH. Interestingly, the HO 2 sensitivity has a reverse dependence on NO levels, decreasing by a factor of B2.3 from the highest NO to the lowest reflecting the decrease in HO x lifetime with increasing NO levels.

ARTICLE IN PRESS
When the model is constrained by both H 2 O 2 and CH 2 O, the increase in OH and HO 2 levels, on average, is 17% less than that which would be estimated from taking the simple sum of the increases calculated independently for H 2 O 2 and CH 2 O. This 17% difference is partially due to the additional loss resulting from the reaction OH + HO 2 ([R7] see Table 1). The contribution from this reaction obviously would rapidly increase with HO x level itself in a quadratic form. For the mid NO level case, there is a factor of B1.5 increase in this loss when comparing model constrained by both H 2 O 2 and CH 2 O with calculation constrained by only H 2 O 2 or CH 2 O. In addition to the loss increase, another important factor is the decrease in the net effect of snow emission on HO x because of the enhancement in HO x . For example, the net contribution from H 2 O 2 to HO x can be evaluated by Eq. (1): where j 6 and k's are reaction rate coefficients. As shown in Eq. (1) It would appear that the thermal decomposition of pernitric acid ([R25]) might be one of the pivotal reactions of this sequence. This reaction can rapidly return HO 2 to the HO x reservoir and yet the rate coefficient for this process, k 25 , has one of the largest uncertainties associated with it at low temperatures (DeMore et al., 1997). This uncertainty has been estimated to be as large as a factor of 10 for SP summertime conditions. Sensitivity tests, however, have suggested that HO x levels are relatively insensitive to the value of k 25 . For a factor of 10 reduction in k 25 , calculated HO x levels decreased by 10% or less. Increasing k 25 by the same factor, however, led to HO x increases of up to 60%. Based on an analysis of NO x and HO 2 NO 2 observations, Slusher et al. (2002) suggested that the upper limit on the uncertainty of k 25 should be no more than a factor of 3.3. In this case, the variation of HO x driven by k 25 would be less than 20%.
Heterogeneous losses of HO x may also be quite important at SP (Chen et al., 2001b). During ISCAT 1998, observations suggested significant HO 2 scavenging frozen fog droplets on two occasions. A similar event occurred over the time period of 12-13 December during ISCAT 2000. At this time, the SP meteorological station recorded nearly 24 h of frozen fog and fog/misty conditions with visibility dropping from over 10 to B1 km. On this occasion, the NO level also dropped from an average of 175 pptv to less than 10 pptv. The photolysis rates jO( 1 D) and j(NO 2 ) decreased by a factor of 1.3 and 1.5, respectively. The levels of CH 2 O reduced from B140 to 35 pptv; while the H 2 O 2 decreased from B250 to B75 pptv. And finally, OH values dropped from an average of 3 Â 10 6 -1.3 Â 10 5 molec cm À3 . Unfortunately, there were no HO 2 values recorded during this time period.
As in ISCAT 1998, we believe that the cited shift in OH during ISCAT 2000 was caused by two factors: (1) a decrease in precursors, i.e., NO and, to a much lesser extent, H 2 O 2 and CH 2 O and (2) scavenging of HO 2 by fog droplets. The impact from dropping NO levels was addressed earlier in this text in Section 3.1 and is directly tied to the recycling of OH through reaction [R4]. Regarding the impact of fog droplets on HO x , there have been field observations showing large HO x decreases in clouds (e.g., Mauldin et al., 1998). Model sensitivity calculations showed a significant reduction in OH due to lower values of NO; however, these values were still nearly a factor of 7 times higher than the observations for the time period corresponding to the on-set of ice-fog. Chen et al. (2001b) previously estimated a median value for heterogeneous loss of HO 2 of 9 Â 10 À3 1 s À1 . This estimate was derived from a fog droplet size range of 5-10 mm having a number density of 5-15 cm À3 . Use of this HO 2 loss rate coefficient led to a decrease by factor of 3 for OH and a factor of 8 for HO 2 . Although the model OH remains to be more than a factor of two higher than the observations, the difference is within the bounds of the combined uncertainty for both model and observations. Even further increase the HO 2 loss rate to the maximum value, the model OH decrease would be less than 20% while HO 2 reduction is proportional to its loss rate increase. This is because the HO 2 is already so low that the contribution from OH secondary formation (HO 2 reaction) is nearly negligible. Although addition of heterogeneous loss to the model has not reproduced the observations of OH, it is evident that the HO 2 heterogeneous loss may be quite important and requires further clarifications. The future study should include insitu simultaneous observations of both OH and HO 2 as well as the number size distribution of SP fog droplets.

Assessment of HO x budget
Fig. 5 summarizes the important SP HO x photochemical processes. As discussed earlier in the text, the snow emissions of H 2 O 2 and CH 2 O must be included as HO x sources not only in Arctic sites (e.g., Hutterli et al., 2001) but also in SP and possibly in other snow covered areas. Considering the sources and sinks shown in the figure, a HO x budget analysis is carried out using the mid NO level (i.e., 89 pptv) condition as defined earlier in Section 3.2. It was noted here that the best agreement between model predictions and observations was found around this NO level (e.g., see Figs. 2a and b).
Summarized in Fig. 6a are the major sources of SP HO x . The largest source is that labeled as 'Snow Emissions'. This source is made up of HO x produced from the photolysis of snow released CH 2 O and H 2 O 2 . They define B32% and 14% of the total HO x source, respectively. Next in line are CH 4 oxidation chemistry and O 3 photolysis. Both of these contribute B27%. Of the total CH 2 O contribution (snow+CH 4 oxidation), 54% is from snow emissions and 46% comes from CH 4 oxidation. Overall, therefore, snow emissions define the single largest source of SP HO x (i.e. 46%). A comparison with the standard model output reveals that snow emissions lead to a factor of 2.3 increase in the total HO x formation rate (vs. a factor 1.8 of in HO x abundance). Similar results have been shown in the Arctic studies carried out in Alert, Canada and Summit Greenland (Yang et al., 2002;Hutterli et al., 2001). Finally, we would like to note that our sensitivity calculations suggest that for the case of zero deposition the snow emission would still make up 23% of the total HO x production. In this case, CH 4 oxidation would be the largest HO x source, contributing 50%; while the remainder 27% is the contribution from the reaction of O( 1 D)+H 2 O.
Major SP HO x sinks are summarized in Fig. 6b. As presented in Section 3.3, the largest of these is the loss of HO 2 through the formation and subsequent deposition of HO 2 NO 2 ([R29]). In addition, HO 2 NO 2 can lead to further HO x loss via its reaction with OH (e.g., [R28]). Together, these two HO 2 NO 2 related HO x loss pathways define B45% of the total. The significant role of HO 2 NO 2 in removing SP HO x reflects the fact that HO 2 NO 2 is reasonably stable due to low temperature and can be rapidly deposited to snow. The radicalradical reaction (i.e., OH+HO 2 , [R9]) contributes another 30%. The final contribution comes from the reaction of OH with NO 2 to produce HNO 3 , [R20], which then mainly undergoes deposition to the surface snow ([R23]). The latter process makes up the last 24% of the total HO x sink. The above cited losses lead to an overall HO x lifetime estimate of 5.8 min. By comparison, the lifetime estimated using the standard model is 6.8 min.
Significant differences exist in the estimated HO x losses for ISCAT 2000vs. ISCAT 1998. Chen et al. (2001a suggested that the major HO x loss (i.e., 62%) was through HNO 3 reactions ([R23] and [R22]). In ISCAT 2000, however, the major HO x loss pathway was shown to be through HO 2 NO 2 reactions. Furthermore, at the lower NO levels encountered in 2000 formation of HO 2 NO 2 ([R13]) is more efficient than that of HNO 3 ([R42]) because of higher HO 2 level. As shown in Figs. 1b and 2b, the HO 2 concentration rapidly decrease with increasing NO levels. A close examination of the model results revealed that the HO x loss through HO 2 NO 2 is more important when NO levels are lower than B120 pptv. On the other hand, at a given NO level, our sensitivity test shows the HO 2 level is a strong function of the HO x primary production (e.g., [R2]). Crawford et al. (2001) has shown strong evidence suggesting surface photochemistry can play a significant role in modifying the SP surface ozone seasonal trend and generate late Spring/early Summer perturbations in O 3 level that exceeded the Winter maximum levels. These authors reported a median model estimated net ozone production rate of 2.8 ppbv day À1 . This production rate is equivalent to nearly 10% of average SP surface ozone concentration per day. Based on the observations from 2000, we have estimated the median net ozone production rate (P(O 3 )) of 4.2 ppbv day À1 or nearly 13% per day. As shown in Fig. 7, the P(O 3 ) is a strong non-linear function of NO level. The peak P(O 3 ) value of B6.5 ppv day À1 situates around 160 pptv of NO and the inner quartile range was estimated to be from 3.2 to 4.8 ppbv day À1 . The higher values in 2000 can mainly be attributed to two factors: observed NO range and extra HO x sources from snow emission of H 2 O 2 and CH 2 O. Recall the observed NO median for the 1998 study was 225 pptv (Davis et al., 2001) and the median derived from observations of 2000 is only 89 pptv. As shown by Crawford et al. (2001), the net ozone production rate has fallen to less than half of the peak value at 225 pptv of NO. By contrast, well over 73% of model runs for 2000 data give P(O 3 ) values larger than the half of the peak value of B6.5 ppbv day À1 . The nonmonotonic relation between P(O 3 ) and NO has been discussed by Crawford et al. (2001). These authors have shown the linkage the intensity of photochemistry with the magnitude of P(O 3 ). This can be also seen here by comparing the Figs. 1 and 2 with Fig. 7. In fact, the P(O 3 ) and model predicted OH is highly correlated with an R 2 value of 0.97. This is because both P(O 3 ) and OH levels are largely controlled by the reaction of HO 2 with NO.

Implications to SP surface ozone
Generally, the P(O 3 ) trend closely mimics that of the gross ozone production rate (F(O 3 )). This reflects the fact that the photochemical ozone destruction rate (D(O 3 )), on average, is less than 18% of F(O 3 ) values. The major ozone production channel is the reaction of HO 2 +NO, contributing nearly 80% of total. The remainder is through the CH 3 O 2 +NO reaction. The two largest O 3 destruction pathways are the reactions of OH+O 3 and HO 2 +O 3 , constituting B46% and B23% of the total, respectively. The relative importance of these two processes can be understood by the low HO 2 / OH ratio of B16 and the enhanced reaction rate coefficient difference of a factor of 33, resulting from the SP environmental conditions characterized by low temperature and high NO levels.  Fig. 6. HO x budget analysis for SP summertime Photochemistry: (a) sources, (b) sinks. This analysis is based on the use of the mid NO case at 89 pptv.
of SP. The remaining 18% of ozone loss is through a sequence of reactions starting from NO+O 3 and followed by NO 2 +OH.

Summary and conclusions
The OH levels recorded in ISCAT 2000 now place the average 24 h oxidizing capacity of near surface air at SP to be larger than that in the tropical marine BL. The current analysis of SP HO x chemistry has provided strong evidence that SP HO x levels not only have a strong dependence on snow emissions of NO but also on emissions of CH 2 O and H 2 O 2 . Both compounds were found to be present at SP at levels much higher than those predicted based on gas phase photochemistry alone. The model analysis suggests that the observations of HONO are inconsistent with both the observations of HO x and NO. The difference between observations and model predications far exceeds the limits defined by the combined uncertainties for the model and observations. By contrast, constraining the model with observed values of CH 2 O and H 2 O 2 leads to model predicted HO x levels that are well within the aforementioned uncertainties. Our results related to HONO would appear to be in conflict with those found at Summit Greenland and at other Artic polar sites where HONO has been identified as the major OH source. On the other hand, it must be recognized that no direct observations of OH were available at these Arctic sites to test their conclusions.
A HO x budget analysis has revealed that snow emissions of CH 2 O and H 2 O 2 may constitute the single largest HO x source at SP. The total contribution is estimated at 46%, 32% of which could be assigned to CH 2 O and 14% to H 2 O 2 . The other important sources are identified as CH 4 oxidation chemistry and O 3 photolysis (i.e., O( 1 D)+H 2 O), each contributing B27%. The greater efficiency of CH 2 O in its influence on HO x levels primarily reflects the five times higher photolysis rate of CH 2 O. Over 93% of the impact from H 2 O 2 is estimated to come from snow emissions of this species. In contrast, only 60% of the impact from CH 2 O is from snow emissions, with the remainder coming from CH 4 oxidation chemistry.
The analysis of SP HO x sinks indicates that the largest sink is through the reaction [R24], HO 2 +NO 2 (45%), followed by HO 2 NO 2 dry deposition and reaction with OH ([R27] and [R26]). The second largest sink involves the radical-radical reaction, OH+HO 2 (e.g., 30% of the total). Loss of HO x through HNO 3 contributes B24%. The relative importance of HO x loss through HO 2 NO 2 vs. HNO 3 is primarily dictated by the ambient NO level.
It is evident that additional polar studies of HO x chemistry will be required. Having available simultaneous observations of HO 2 and OH will be a critical need. Obviously, a resolution of the importance of HONO as a source of OH and NO x will also be critical. The latter will require new specific measurement techniques for HONO as well as a reassessment of model HO x sink processes.