The potential importance of iodine for marine boundary layer (MBL) chemistry has found increasing recognition in recent years. However, observations of the key iodine species are sparse and the chemical reactions of the iodine oxides are not well understood. Here we present Differential Optical Absorption Spectroscopy observations of IO, OIO in the MBL of the Gulf of Maine, U.S., during Summer 2004. We report the first daytime observation of OIO, indicating that this compound is rather photostable. Mixing ratios of IO were similar to, and those of OIO higher than, values reported for European coastal sites. Calculations with the one-dimensional model MISTRA show that the observed simultaneous presence of elevated OIO and NOx cannot be explained by currently known iodine chemistry. Our results lead to the conclusion that thus far unknown chemical reactions of iodine oxides, probably involving iodine nitrates, might occur in the MBL.
 The chemistry of reactive iodine species (RIS) in the mid-latitude marine boundary layer, MBL, has received considerable attention since the first observation of elevated levels of IO at Mace Head, Ireland [Alicke et al., 1999]. RIS can significantly influence ozone chemistry, oxidation capacity, and particle formation and growth in the MBL. It is currently believed that RIS originate from algae which emit iodo-methanes and I2 [e.g., Saiz-Lopez and Plane, 2004]. These compounds are then rapidly photolyzed, releasing iodine atoms which further react with ozone to form IO. IO can undergo a number of reactions in the MBL. The reaction with NO2 leads to the formation of IONO2, which can thermally decompose, photolyze, or be taken up into the aerosol. IO can also react with itself or BrO, forming OIO with yields of 40% and 80%, respectively [Bloss et al., 2001]. Thus far, atmospheric OIO has been observed only during the night [Allan et al., 2001; Saiz-Lopez and Plane, 2004; Saiz-Lopez et al., 2006]. Since the levels of the OIO precursor IO are highest during the day, Allan et al.  and Saiz-Lopez et al.  concluded that a fast removal process for OIO must be present during the day. Because the upper limit photolysis yield is 0.1 [Joseph et al., 2005; Tucceri et al., 2006] photolysis might be important under low NOx conditions. Under conditions of elevated NOx, the reaction OIO + NO → IO + NO2, could explain the lack of daytime OIO [Plane et al., 2006]. Similarly, the OH + OIO reaction could significantly contribute to the loss of OIO [Plane et al., 2006]. Other OIO loss mechanisms proceed through the formation of new particles, as well as possible intermediate products (such as I2O3). Despite a number of laboratory studies, these mechanisms are still not well quantified and there is a high likelihood that other presently unknown OIO reactions are important in the MBL. Consequently, further atmospheric observations are necessary to test our understanding of IO and OIO chemistry. Here we present the first observation of daytime OIO together with measurements of IO in the Gulf of Maine. The observations are compared with model calculations to elucidate the chemistry of reactive iodine compounds in the coastal polluted MBL.
2. Experimental Method
 A collaborative field experiment on marine boundary layer Chemistry of Halogens at the Isles of Shoals (CHAiOS) was conducted from July 4–August 10, 2004 as part of the ICARTT study. The Isles of Shoals are a group of twelve small islands dispersed in an area of ∼4 × 4 km, that are located in the Gulf of Maine, 10 km off the coast near Portsmouth, New Hampshire, USA (42°58′ N–70°37′ W). The island shores are populated with kelp, including codium fragile, desmarestia aculeate, agarum clathratum, and laminaria spp. [Harris and Tyrrell, 2001] Some of the kelp species are restricted to the shores due to the rapidly increasing ocean depth.
 Halogen oxides, ozone, and NO2 were measured by Differential Optical Absorption Spectrocopy (DOAS) [Stutz and Platt, 1996]. The absorption path of UCLA's Long Path (LP) instrument ran from Appledore Island to White Island 35 −15 m above the open ocean surface with a total length of 4.6 km. The shores of other islands were within a few hundred meters of the light path. The Multi-Axis (MAX) DOAS instrument on Appledore Island sequentially collected scattered sunlight from five elevation viewing angles: α = 1°, 3°, 5°, 15°, and 90°, with an azimuth viewing angle almost parallel to the LP DOAS line of sight. Photolysis rates were determined with a Bentham/Gigahertz Optics spectroradiometer and literature absorption cross-sections (for example Ashworth et al. , Hönninger ).
 LP-DOAS spectra were analyzed as previously described [Stutz and Platt, 1996; Stutz et al., 2002] in the following spectral intervals (literature absorption cross section in parenthesis): 322.1–371.9 nm for NO2 (Voigt et al. ), 324.1–345.6 nm for O3 (Voigt et al. ), 423.6–447.7 nm for IO (Hönninger ), 541.9–564.5 nm for OIO (P. Spietz, Univ. Bremen, personal communication, 2005) and I2 (Saiz-Lopez et al. ). MAX-DOAS spectra were analyzed from 414.5–439 nm for IO, and 541–566.7 for OIO and I2. In addition to the trace gas absorption cross-sections listed above, a temporally close zenith spectrum, and a Ring spectrum were included in the spectral analysis of the low elevation angle MAX-DOAS spectra. Details on the spectral analysis of OIO are given in the auxiliary material. Both LP- and MAX-DOAS instruments clearly identified daytime OIO at the Isles of Shoals (Figure 1).
 The analysis of MAX-DOAS absorption spectra results in differential slant column densities (DSCD), i.e. the difference between the trace gas columns measured with a low-elevation angle viewing direction, α, and those measured aiming at the zenith. For trace gases horizontally and vertically evenly distributed in the MBL, DSCD are, in a first approximation, proportional to 1/sin(α), i.e. DSCD's for low elevations are higher than for higher elevations. We adopted a more precise approach to convert DSCD's into vertical column densities and concentrations by using a Monte-Carlo radiative transfer model (TRACY), which we constrained by meteorological observations and O4 DSCD's as described by [Hönninger et al., 2004; Pikelnaya et al., 2007]. OIO was assumed to be evenly distributed in the lowest 100 m of the MBL. OIO levels were assumed to be zero above 100m. The reported error of the MAX-DOAS mixing ratio includes both, the uncertainty of the fitting procedure (see auxiliary material) and the error of the radiative transfer calculation [Pikelnaya et al., 2007].
3. Model Description
 To interpret our observations we used the one-dimensional MBL model MISTRA, which describes MBL chemistry in the gas and aerosol phase, as well as aerosol microphysics [von Glasow et al., 2002]. The chemistry scheme contains state of the art iodine chemistry [see Pechtl et al., 2006], with updated reaction rate constants for OIO + NO → IO + NO2 and OIO + OH → HIO3 according to Plane et al. . It also includes a parameterization for new particle formation related to iodine oxides [Pechtl et al., 2006]. In the model, OIO is produced via IO self-reaction and reaction of IO with BrO. OIO sinks are the reactions with OH and NO, photolysis, uptake by pre-existing particles, and new particle formation (see supplement of Pechtl et al. ).
 In order to compare our model results with the Isles of Shoals data, we performed model runs for a moderately polluted atmosphere with 2 nmol mol−1 of NO2 and a polluted scenario with 7 nmol mol−1 of NO2. For both scenarios, we present a base case, where OIO is assumed to be photolytically stable and to form new particles, and sensitivity studies where we include OIO photolysis (with a quantum yield of 5%) or exclude particle nucleation. The model was used in a Lagrangian mode in which the model column is assumed to move first from the continent over the oceans and reach the Isles of Shoals after 1 hour. Emissions of iodine precursors occur only along the coastlines of the different islands (“hot spots”) near the LP-DOAS light path. Six emission “hot spots” of organoiodides and I2, each lasting 10 seconds, are encountered by the air column at intervals of 2 minutes. All scenarios are run for 3 hours starting at local noon. Please note that the prescribed surface fluxes of organoiodides and I2 are not well constrained from observations, resulting in large uncertainties in quantitative iodine oxide mixing ratios. Therefore we will focus in the discussion of the model results on a comparison between the different scenarios.
 The reactive iodine species IO and OIO were observed on 19 out of 23 days during the CHAiOS experiment. Figure 2 shows a three day period which is representative of our RIS observations during the entire experiment. IO was observed only during daytime, with maximum levels of 4 pmol mol−1 (Figure 2). OIO was detected with mixing ratios of ∼10 pmol mol−1 and a maximum of ∼30 pmol mol−1. In contrast to IO, OIO levels were comparable during day and night (Figure 2).
 The observations show a number of features that previously have not been reported. First and foremost, OIO was observed for the first time during the day (see also Figure 1). This observation is in contrast to those made by Saiz-Lopez et al.  at Mace Head, Ireland where OIO was only seen at night. The lack of a tidal signature in the IO mixing ratios, which was observed at other locations [Saiz-Lopez and Plane, 2004], is currently unexplained.
 MAX-DOAS OIO DSCD's for 1°, 3°, 5°, and 15° elevation angles (Figure 2) were separated, as expected for a fairly spatially homogeneous trace gas near the surface. The OIO mixing ratios, calculated for vertically evenly distributed OIO in the lowest 100 m of the MBL, agree well with the LP-DOAS data (lower panel in Figure 2), considering the uncertainties in both measurements. In contrast, IO DSCD's (not shown here), which followed the temporal trend of the LP-DOAS observations, did not show a separation between the elevation angles. We believe the lack of separation is caused by the presence of IO in small scale plumes (see also Saiz-Lopez et al. ), of which we may have probed only one. The clear DSCD separation observed for OIO shows that the spatial distribution of OIO is quite different from that of IO. Based on the MAX-DOAS viewing geometry and the good agreement between MAX-DOAS and LP-DOAS OIO mixing ratios, we believe that OIO was distributed more evenly than IO in the horizontal, probably on scales of up to a few kilometers, and in the vertical on the scale of 100 m or more. To be spatially more evenly distributed, OIO must be longer lived than IO in the MBL. The details of the MAX-DOAS observations and their interpretation are discussed by Pikelnaya et al. .
 Our observations also do not show an anticorrelation between the iodine oxides and NO2, which is expected from known chemistry, i.e. the reaction of IO with NO2 and OIO with NO. This observation becomes particularly clear on August 5. In the morning, 5–15 pmol mol−1 of OIO and 1–4 pmol mol−1 of IO were observed in the presence of 5–10 nmol mol−1 of NO2. In the afternoon, the IO and OIO levels remained similar while NO2 mixing ratios were much lower.
 To test whether these observations can be explained by our current understanding of iodine chemistry, we discuss the temporal evolution of iodine oxides and NOx for the semi-polluted and polluted model scenarios at an altitude of 25 m (Figures 3 and 4), which resembles most closely the altitude of the LP-DOAS light path. For the base cases, OIO peak mixing ratios are more than one order of magnitude lower in the polluted case than in the semi-polluted case, mainly due to the sink reaction OIO + NO → IO + NO2. This reaction produces IO, partly balancing the loss of IO by the formation of IONO2. Therefore, in the model IO shows a much weaker dependence on NOx levels than OIO. IO/OIO ratios are ∼2 for the semi-polluted and ∼40 for the polluted base case. This is in clear contrast to our observations, where OIO mixing ratios tend to exceed IO mixing ratios, and IO/OIO ratios do not exhibit a dependence on NOx concentrations.
 If nucleation of OIO is excluded, modeled OIO mixing ratios strongly increase in the semi-polluted case, but not in the polluted case, where the reaction with NO is the dominant sink for OIO. This shows that a potential overestimation of the nucleation sink for OIO cannot explain the lack of agreement between model results and measurements. If possible intermediate products between OIO and particles are included (I2Oy after Saunders and Plane ), the model predicts OIO mixing ratios decreasing to below the instrumental detection limit in all cases (not shown). Including moderate OIO photolysis in the model (see also discussion by Pechtl et al. ) does not change the results significantly.
 The modeled vertical profiles of IO and OIO (not shown) do not show a more spatially uniform distribution of OIO than IO. This is in contrast to our interpretation of the MAX-DOAS observations. According to the present understanding of iodine chemistry, OIO cannot survive in a NOx-rich environment; i.e., even if background sources exist and even if it is photolytically absolutely stable, one would not expect it to be transported far.
 The model results show that our current understanding of iodine chemistry, as implemented in MISTRA, is not able to explain the observations on Appledore Island. A possible way to reconcile the observations with the model would be the inclusion of thus far unknown reaction pathways that convert iodine reservoir species back into OIO. The reaction IO + NO3 → OIO + NO2, suggested by Saiz-Lopez et al.  does not produce significant levels of OIO during daytime in the model. In the model, IONO2 contains about 35% of gas-phase iodine in the semi-polluted base case and about 80% in the polluted base case. If this reservoir gas could be recycled, for example with a reaction like IONO2 + O3 → OIO + NO2 + O2 (even though this reaction seems unlikely to occur as pure gas phase reaction, J. Crowley, personal communication, 2005), IONO2 could be partly converted back to OIO. We want to stress that a multi-step reaction possibly involving aerosol surfaces would be an alternative for a single-step, pure-gas phase reaction as we implemented it here. The inclusion of this reaction in the model with a hypothesized reaction rate coefficient of 10−13 cm3molec−1s−1 leads to an increase in OIO mixing ratios, especially in the polluted case, where IONO2 levels are higher (see green, dash-dotted lines in Figures 3 and 4). IO and OIO values are now similar in both cases, leading to a better agreement with the observed independence of iodine oxide concentrations from NOx levels. IO/OIO ratios decrease compared to the base case for high and low NOx conditions, which also improves the agreement with our observations. However, the suggested reaction does not increase the atmospheric lifetime of OIO compared to IO, because the primary source for the key species IONO2 is the reaction of IO with NO2. Hence, the observations of the MAX-DOAS instrument of spatially more homogeneously distributed OIO can still not be reproduced in the model. To explain the MAX-DOAS observations, unknown cycles of OIO that do not involve IO as intermediate product seem to be necessary or the measured rate coefficient for the reaction OIO + NO is much lower than has been measured which we, however, regard as unlikely.
 Elevated levels of IO and OIO were observed by two separate DOAS instruments in the MBL in the Gulf of Maine. Daytime OIO was observed for the first time, indicating that this compound is relatively photostable. A comparison of the reactive iodine levels discussed here with those previously reported at Mace Head [Alicke et al., 1999; Saiz-Lopez et al., 2006] show similar IO but higher OIO levels in the polluted Gulf of Maine. The model results predict – in contrast to our field observations–significantly higher OIO levels in a clean environment than in the presence of high NOx. Unknown reactions recycling OIO from IONO2, the largest iodine reservoir under polluted conditions, would be a possibility to explain our observations. The observation that OIO is spatially more evenly distributed than IO can be explained by a longer chemical lifetime of OIO, e.g. unknown rapid OIO recycling processes. The comparison of our observations with the model results illustrates that our understanding of reactive iodine chemistry in the MBL still has considerable gaps.
 We would like to thank the CHAiOS team and the staff at the Shoals Marine Laboratory for their logistical support. Funding for this study was provided by the National Science Foundation (ATM-0401599), the NOAA Health of the Atmosphere Program (RA133R-04-SE-0411), and the DFG - Emmy Noether Program (grant GL 353/1-1,2). Additional travel funds were provided by the NSF (grant ATM04-01611). This paper is contribution number 147 to the Shoals Marine Laboratory.