Ozone depletion due to dust release of iodine in the free troposphere.

Iodine is an atmospheric trace element emitted from oceans that efficiently destroys ozone (O3). Low O3 in airborne dust layers is frequently observed but poorly understood. We show that dust is a source of gas-phase iodine, indicated by aircraft observations of iodine monoxide (IO) radicals inside lofted dust layers from the Atacama and Sechura Deserts that are up to a factor of 10 enhanced over background. Gas-phase iodine photochemistry, commensurate with observed IO, is needed to explain the low O3 inside these dust layers (below 15 ppbv; up to 75% depleted). The added dust iodine can explain decreases in O3 of 8% regionally and affects surface air quality. Our data suggest that iodate reduction to form volatile iodine species is a missing process in the geochemical iodine cycle and presents an unrecognized aeolian source of iodine. Atmospheric iodine has tripled since 1950 and affects ozone layer recovery and particle formation.

INTRODUCTION

Iodine is a critical micronutrient for human health transported through, and possibly partially acquired from, the atmosphere (, ). Atmospheric iodine is prevalent in the marine boundary layer (MBL) (), lower free troposphere (), upper free troposphere (), and stratosphere () and participates in rapid photochemical cycles, which destroy ozone (O3) and modify the atmospheric oxidative capacity (, –). Iodine photochemistry decreases the tropospheric O3 burden by 9% (, ), decreasing radiative forcing (, ), and reduces the atmospheric OH burden, increasing the lifetime of methane and other greenhouse gases (). On a per-atom basis, iodine is about three orders of magnitude more efficient than chlorine at destroying O3 (). Iodine oxoacids nucleate particles more efficiently than sulfuric acid (), and can dominate new particle formation (NPF) and particle growth in coastal areas with macroalgae (, ) and in the Arctic ().

The global source of iodine is dominated by O3-stimulated emissions of volatile inorganic iodine species [i.e., hypoiodous acid (HOI) and I2) from the ocean surface (~1.86 Tg year−1) (, , ). Since 1950, anthropogenic O3 has increased the northern hemisphere iodine burden threefold over preindustrial levels via this stimulated emission (–). The oceans also dominate the atmospheric sources of organic iodine compounds, primarily methyl iodide (CH3I), via biotic and abiotic processes (–). Biogenic sources of iodine—both marine and terrestrial—arise from iodine’s role as a micronutrient (, ). Terrestrial iodocarbon sources are much smaller than marine sources (~0.091 Tg year−1) (, ). However, terrestrial inorganic sources of iodine originated from arid regions have not been previously considered.

Dust layers are often depleted in O3 (–), but the necessary O3 sink within dust plumes remains poorly understood. Previous studies have attributed dust-related O3 loss to heterogeneous loss of O3 and NO species on the dust surfaces. However, laboratory studies have now firmly established that the reactive uptake of O3 is low (reactive uptake coefficient, γO3 ≤ 6.6 × 10−7) (), leaving most of the necessary O3 sink associated with dust unexplained. Previous field measurements have also found large enhancements in gas-phase iodine, both iodine monoxide (IO) and CH3I, associated with dust layers from the Sahara (, ). However, a possible role of iodine in the O3 depletion has not been previously assessed, nor has an aeolian iodine source to the atmosphere been established.

RESULTS

Decoupled layers enriched in iodine and dust

During aircraft flights out of Antofagasta, Chile, in the southern hemisphere tropics, we have consistently detected widespread elevated layers of dust containing enhanced mixing ratios of IO radicals. These layers were detected on the top of the MBL and extended several kilometers into the lower free troposphere (1.5- to 7-km altitude). The dust layers were tracked by the High Spectral Resolution Lidar (HSRL) above and below the Gulfstream V (GV) aircraft (Fig. 1, A and B) to extend from 35.0°S to 12.6°S and up to 455 km from shore and persisted for the entire period (13 days) of aircraft observation. The Airborne Multi-Axis Differential Optical Absorption Spectroscopy (AMAX-DOAS) instrument found IO enhancements for all 27 layer intercepts when clouds did not preclude observation (table S1). This included multiple intercepts of likely contiguous layers along and away from the coast, indicating that IO enhancements are a persistent feature of these layers (fig. S1).

Fig. 1.

Overview of observations.

(A) Sources and extent of observed layers. Flight tracks for TORERO RF01 to RF06 are shown in blue where layers are contiguous with observations of enhanced IO and/or HSRL linear depolarization shows that layers below 6 km are significantly nonspherical (indicates dust); otherwise they are shown in gray. Letter labels indicate heading changes and connect with fig. S1. The background shading (yellow to red) represents the sources of observed dust based on 72-hour back-trajectories (see the Supplementary Materials for details). (B) DOAS and HSRL data for the takeoff of RF05 (RF05-01); the time from takeoff connects with the inset in (A). (C) IO spectral proofs from the case study shown in the center panel. Individual spectral proofs are vertically displaced by 1 × 10−3 in optical density for separation; numbers on the left indicate the retrieved IO mixing ratio with 1σ uncertainties, and those on the right in square brackets indicate the altitude of the observation.

Figure 1C shows spectral proof of IO fingerprint absorption, and its variation with altitude during one of these dust intercept vertical profiles. The signals inside the lofted dust layers exceed those found in the MBL. The observed IO mixing ratios above, below, and between the dust layers are lower and found consistent with regional free tropospheric background of 0.10- to 0.25-pptv (parts per trillion by volume) IO (). Within the dust layers, the observed IO is as much as a factor of 10 higher than this background. DOAS observations of oxygen-oxygen collision-induced observation (O4; a reliable atmospheric standard) indicate that the dust layers reduce path length in aggregate rather than enhance it by multiple scattering, establishing unambiguously that IO concentrations are enhanced in the dust layers (fig. S2; see the Supplementary Materials for details). The IO enhancements are consistently observed in dust layers at low and high altitudes.

Back-trajectories initiated along the flight track show no contact with the MBL over up to 10 days and instead point to a terrestrial source from the Atacama, Monte, and Sechura Deserts. These deserts are portions of the South American Arid Diagonal (SAAD) (). Figure 1A shows that the Atacama is the dominant source region for the intercepted layers (at least partially a reflection of proximity to most flight tracks). However, data from three layer intercepts near Peru are unambiguously from the Sechura Desert and not from the Atacama. The Atacama is the site of a globally unique extraction of iodine from caliche deposits (). However, these deposits are located below the surface. Furthermore, while ore-processing facilities are located in the vicinity of many back-trajectories (i.e., located near Antofagasta where the aircraft was based), most of the back-trajectories from dust layers do not pass within 0.5° latitude or longitude of these facilities and many travel well above the boundary layer. We conclude that dust and iodine arise from a common source, i.e., the various deserts spanning the SAAD.

Ozone depletion due to iodine in the free troposphere

The lofted dust layers are consistently accompanied by significant O3 depletion. Figure 2 shows an example of these ozone-depleted laminae in the free troposphere. Potential temperature and humidity discontinuities at the layers’ edges indicate that the vertical extent of the layers is dynamically controlled. As noted above, the layers extend horizontally over hundreds of kilometers with average O3 depletion of 17% but as much as 40% (table S1). Near the layers’ cores, O3 drops exceptionally low [O3 ≤ 10 parts per billion by volume (ppbv) for 9 of 22 layers with O3 measurements; table S1], creating ozone-depleted laminae. Aerosol counters show distinct submicrometer and supermicrometer aerosol populations in all three layers (even above 5 km; fig. S3), with the larger aerosol gravitationally settling within each layer, as indicated by the mean aerosol diameters (Fig. 2, middle). HSRL depolarization data (fig. S1) show that layers below ~5 km are nonspherical dust but become spherical at high altitude and with increasing distance from the coast (see the Supplementary Materials for details). Across all three layers, the degree of O3 depletion correlates with aerosol surface area, consistent with the hypothesis of heterogeneous O3 uptake—however, so does the IO mixing ratio (Fig. 2, right).

Fig. 2.

Summary of RF05-01 case study.

Left: Relative humidity and potential temperature. Middle: Demonstration of the anticorrelation of O3 and aerosol surface area consistent with the literature and previously attributed to heterogeneous uptake. Gray shading across panels indicates when this smoothed surface area is greater than 25 μm2 cm−3; uppercase letters indicate the individual dust-layer intercepts. Dashed lines are indicative of a contiguous background. Right: Iodine observations; error bars show 1σ errors. Mixing ratios of IO are retrieved by the AMAX-DOAS using a parameterization method. Dashed line shows median of southern hemisphere TORERO observation after filtering for dust-impacted data. Aerosol volume density (box-car smoothed over 15 s for sufficient statistics) is shown. Water vapor was measured by the Vertical Cavity Surface-Emitting Laser (VCSEL) hygrometer. CH3I is measured with Trace Organic Gas Analyzer (TOGA).

Recent laboratory studies have established that O3 reactive uptake onto dust proceeds too slowly under atmospheric pressure to contribute substantially to O3 loss () due to most O3 regenerating when O2 is present (see the Supplementary Materials for details). Heterogeneous uptake of NO also contributes to O3 destruction. However, as highlighted by DOAS NO2 measurements (fig. S2), the observed layers exist in a low NO environment, which limits the potential impact of heterogeneous NO uptake in this study. We examined the capacity for heterogeneous uptake to cause the observed O3 depletion in the laminae, confirming a negligible impact from direct O3 uptake and modest impacts (ΔO3 ~ 10 ppbv after 7 days) from reactive NO uptake (fig. S4). Heterogeneous uptake of NO is important but has two major limitations: (i) The amount of odd oxygen directly removed is small (NO << O3), and (ii) removal of NO suppresses O3 production but does not produce persistent chemical cycles to consume O3. Even using theoretical maximal values for NO uptake (γ = 1), the impacts on ozone are the same (<0.1 ppb O3 difference at all times). As a result, O3 mixing ratios stabilize above 10 ppbv. Known heterogeneous reactions of O3 and NO species are not enough to explain the observed low O3 mixing ratios in the ozone-depleted laminae.

At the observed IO levels, gas-phase iodine chemistry is highly efficient at destroying additional O3 through catalytic photochemical reaction cycles. Introducing 10 pptv of Iy (consistent with ~2 pptv of IO observed; fig. S4) accelerates O3 destruction by over a factor of 2 and is needed to explain the low observed O3 at the core of these layers. Only when the iodine is included can the box model explain the low O3 below 10 ppbv (fig. S4). Iodine directly destroys O3, dominates O3 destruction as shown by the reaction integrals over 3 days (tables S4 and S5), and remains efficient even at these low O3 concentrations. Critically, a distinct anticorrelation is observed: Iodine levels are highest where O3 is lowest (fig. S5). Vigorous iodine chemistry consistent with known chemical mechanisms of gas-phase O3 destruction by observed IO levels is needed and found to be sufficient to explain the ozone-depleted laminae.

It is possible that the multiphase reaction I− + O3 enhances the ozone sink beyond the gas-phase iodine chemistry modeled here. Any O3 loss due to multiphase chemistry depends on the gas-phase O3 concentration and on how the conditions sustaining I− in aerosol change inside the dust layers relative to the background. Compared to the stratosphere (), the contribution from I− + O3 in dust layers is masked by rapid gas-phase chemistry at much (factor ~ 150) higher gas-phase IO radical concentrations, lower O3 concentrations (factor ~ 0.1), and the dilution of any dissolved iodide concentration in more abundant liquid water inside dust layers. Iodide in dust aerosol is not enriched over background marine aerosol (). Moreover, iodide would be depleted in less than 1 s if not replenished from other iodine reservoir species [e.g., iodate photoreduction (, ); see also the Supplementary Materials]. The photosensitized reduction of iodate in chromophoric dust proxies, while plausible as a source of I− as an initial intermediate toward activating Iy,gas, is currently insufficiently constrained experimentally to model any accompanying multiphase O3 loss with certainty. However, iodide destroys O3 in a stoichiometric ratio of one, meaning that if 10 pptv Iy,gas is liberated via I− + O3, this reaction consumes only 10 pptv of O3. The direct O3 sink of multiphase chemistry to liberate Iy,gas is therefore very small compared to efficient gas-phase chemistry at the high observed IO radical concentrations, which destroys several ppbv of O3. However, in the absence of a gas-phase mechanism to form HIO3 (and thus particulate iodate), O3 loss could be further enhanced and is estimated conservatively here. In the future, more measurements to understand iodine speciation and phase partitioning are needed to understand iodate as a missing component of the geochemical iodine cycle in models.

Aeolian iodine source from dust

While the large IO enhancements inside dust layers could, in principle, reflect a change in chemical partitioning of gas-phase iodine, we find that the dust itself is the source of the iodine. This is because, even under exceptionally low O3 inside the dust layers, the IO/Iy ratio remains high, and IO accounts for at least 15% of daytime Iy. The photostationary equilibrium of gas-phase Iy reservoir species is limited by HOI photolysis and establishes within minutes (daytime). Therefore, changes in gas-phase chemical partitioning induced by dust are small, and to explain elevated IO inside dust layers, an additional iodine source from the dust is needed. We posit that iodate reduction and volatilization is the source of iodine from dust.

We propose that the observed correlation of iodine with aerosol volume (fig. S5) results from sparingly soluble iodates on dust surfaces, which accumulate from atmospheric deposition and geogenic deposits (, –) before lofting. We suggest that acid deposition from the atmosphere onto dust, dust deliquescence, and possibly illumination may be critical steps facilitating iodine release from lofted dust (see Fig. 3 and below): the alkalinity of dust, scarcity of water, and limited illumination before lofting trap iodate. We attribute the sphericity of particles in the layers at high altitude and with increasing distance from the coast to dust deliquescence. This role of water vapor in the activation process is consistent with previous findings that water vapor correlates with the level of O3 depletion associated with Saharan dust (). Liquid water helps to mobilize sparingly soluble IO3− and also iron ions, which facilitates the photochemical production of peroxides either directly () or by using the iron ions (). Iodate might then be reduced through reaction with the iron ions (, ), by H2O2 (), by nitrite () photosensitized reactions (, ), or by numerous other species (). Iodate reduction is thermodynamically favored over nitrate reduction and might serve to suppress renoxification. Acid uptake to lower pH, possibly aided by photosensitized reactions, may be necessary for iodate reduction to proceed at an atmospherically relevant rate. At higher altitudes, dust has an important role as ice nuclei (), and iodine, in turn, recycles on ice surfaces in the upper troposphere (Fig. 3) (, , ). Iodate has been previously observed on Saharan dust during several ship cruises and is notably depleted for the smallest dust aerosol sizes, which is consistent with surface area–dependent heterogeneous release (fig. S6) (, ). Iodate is present in Saharan dust [and caliche deposits below the Atacama; see (, ) and the Supplementary Materials] and is the likely source of iodine enhancements associated with dust that had previously been observed over the Atlantic (, ).

Fig. 3.

Schematic of dust release of iodine and O3 loss.

Broadly clockwise from bottom left: (1) Aeolian processes loft dust, leading to size sorting due to gravitational settling. (2) The basicity of dust will encourage uptake of gas-phase acids such as nitric acid (HNO3) and other NO species. (3) Once deliquesced aqueous ion chemistry is facilitated, iodate reduction can be facilitated by peroxides or soluble iron, either of which would suffice. The redox chemistry of iodine is highly complex, and the precise mechanism leading to volatilization is unknown. Iy is likely to be liberated in the oxidation state of I0 (I2) or I+ (HOI, ICl, IBr) or as an organic species (e.g., CH3I), although the latter is not observed in TORERO. (4) Released iodine photolyzes initiating rapid photochemical cycles. (5) Dust can participate in low-altitude cloud formation, is subject to wet deposition, and can nucleate ice, facilitating recycling of iodine on ice surfaces.

In contrast to the observations from the Saharan region, we do not observe any impact from dust on CH3I (Fig. 2, right). This suggests that there is likely to be more than one mechanism of iodate reduction leading to possible variations in the speciation of volatile iodine released from dust. Laboratory studies of iodate reduction under conditions that resemble atmospheric aerosols are needed to develop missing mechanisms of iodate reduction in atmospheric models.

Atmospheric implications

The precise mechanism of iodine release from dust is currently unknown. However, we approximate the net effect of iodate processing (Fig. 3) in the CAM-chem model (, ) as an autocatalytic dust source, tuned to resemble observed dust-impacted IO concentrations in the study area. The model captures the IO observed in the absence of dust, as well as the average IO enhancements in the presence of dust, yet it exhibits greater variability when dust is present (fig. S7). However, the latter may be expected given the limited number of observations. The spatial scale of the DOAS measurement is likely to include components of background air with lower IO. Nonetheless, IO enhancements of up to a factor of 10 over background were observed within the dust layers. The model underestimates the magnitude and variability of dust-free IO by a small amount in absolute terms, but this leads to a larger and more variable relative enhancement in dust-impacted data. The model has been tuned to capture the average IO concentrations measured in the presence of dust (fig. S7) within the studied region and is not affected by a similar difference in spatial scales. Hence, the median enhancements in fig. S7 are not directly comparable between model and observations, and the observations are expected to be a lower bound on the average IO enhancement inside dust layers. The model accurately produces the approximate vertical extent of the dust layers (Fig. 4B); an extended horizontal domain is chosen to assess atmospheric impacts (allows the displacement of IO enhancements further from dust sources in the model; Fig. 4A). Dust is found to be the source of more than 90% of iodine outflow from the Atacama between 2- and 4-km height and dominates O3 destruction in January (Fig. 4C).

Fig. 4.

Regional (and global) impacts.

(A) Map of IO at 2.9-km altitude in CAM-chem for January 14 (dynamics in CAM-chem are fixed but do not reflect any specific real day). (B) Curtain of IO at 90°W (gray dashed line on map). (C) Iodine release at altitude and comparison of iodine release from dust and from very short lived species (VSLS) computed for January over the box indicated on the maps in Mg month−1 (top axis). Dashed lines show the same traces for an annual extrapolar average; fluxes are shown as Gg year−1 (top axis). (D) Impact of dust iodine on O3 in CAM-chem computed as (O3,dust iodine − O3,base)/O3,base. (E) Curtain of the O3 impact at 90°W. (F) O3 impacts at different temporal and spatial scales. The O3 impact is computed by calculating the mean profiles for the two model cases and then computing the equation above. Displayed values are means computed for the box on the maps for January 14; the months of January and June; for the tropics (defined here as 20°S to 20°N) and for extrapolar regions (defined as 60°S to 60°N).

Ozone impacts due to dust are conservatively estimated in the model and limited to iodine, because NO uptake to dust, as well as dust source of the Sechura Desert, is not represented in CAM-chem. The removal of NO (and by extension NO) will continue to suppress O3 production until replenished, likely increasing its relevance at regional scales. The transport and chemistry of background O3 levels are reasonably well represented (–). The model does not produce conditions similar to the O3-depleted laminae aloft without introducing the new iodine source from dust, which suggests that transport of low O3 air cannot explain the observed O3 depletion in the dust laminae. As expected, modeled localized O3 decreases by ~35% (Fig. 4D), which is less than observed. Nonetheless, a regional (box on Fig. 4, A and D) impact is clearly visible across most of the model domain, lowering O3 by as much as −9.4% (−2.5 ppbv of O3; Fig. 4F). In the free troposphere, iodine destroys O3 150% as efficiently as in the MBL due to increased photolysis rates and has three to six times the residence time, greatly enhancing its impact (fig. S8). Downward transport entrains air into the MBL that is both enriched in iodine (Fig. 4B) and depleted in O3 (Fig. 4E), leading to −6.9% (0.9 ppbv; regional) impact on O3 at the surface and even −4.7% (−0.5 ppbv; regional) in the January mean. The entrainment of iodine-rich and O3-depleted air into the MBL is of relevance to surface air quality and human health, and due to the iodine’s buffer capacity over ozone, pollution is expected to evolve differently within regional scales in the near future depending on the air pollution mitigation policies implemented by different countries ().

On a global scale, the Sahara dominates dust emissions, and the impacts of dust transport on surface air quality have been well documented in cities across Europe (, ). In our simulations, the iodine source from the Sahara is actively controlled and conservatively estimated by modulating the efficiency of iodine release zonally not to exceed the observed levels of IO radicals in the SAAD domain. Consistent with the SAAD dust, Saharan dust has also been observed to affect O3 in the free troposphere at the Monte Cimone Climate observatory, with implications for surface air quality in cities of northern Italy [Po Valley; ()] and southern Spain (). This pattern of dust transport is captured by CAM-chem; however, the magnitude of episodic impacts on O3 of >40% attributed to Saharan dust at Monte Cimone is underestimated in the model, reflecting the conservative implementation of iodine chemistry from Saharan dust (fig. S9). Transport and NO uptake are also likely to contribute substantially to O3 depletion under more polluted conditions such as those observed at Monte Cimone (). Similarly, dust from the Taklimakan and Gobi Deserts have been observed to affect O3 in cities in northern China, although the impacts appear to be smaller (~10%) deep in the continental interior (). While it is very likely that iodine impacts from dust are not limited to the study area, a confirmation of dust iodine release from other deserts globally, including the Sahara and Gobi Deserts, warrants further investigation. In our current conservative implementation, dust is found to be responsible for 41% of iodine release in the extrapolar lower free troposphere, resulting in an annual mean decrease of extrapolar tropospheric O3 by 0.87%, which extends beyond the regional scale.

DISCUSSION

The persistence of IO radicals in aged stratospheric air further indicates that iodate heterogeneously recycles from stratospheric aerosols (), which is consistent with our findings for dust. In the troposphere, particulate iodine is only a minor reservoir of total inorganic iodine (sum of gas- and particulate-phase Iy); most tropospheric iodine resides in the gas phase (). The observations presented here suggest that iodate can be a primary source of iodine to the atmosphere. Together with the rapid oxidation and volatilization of iodide, this points to multiphase iodine chemistry being extremely dynamic and context dependent and warrants further research.

Surface deposits of iodine are the result of atmospheric deposition and geogenic iodate deposits in the Caliche layers of the Atacama Desert. Field measurements indicate that atmospheric HIO3 is widespread () and available to deposit and accumulate on alkaline dust as iodate before lofting. In principle, also iodides, if present, can be oxidized to iodates in an oxic atmosphere (). Iodate in the subsurface Caliche layers is thought to be Jurassic marine deposits uplifted by the Andes orogeny, and slowly transported to their current location by ground water (, , ). Marine diagenesis is the major source of such geological iodine, given that the crustal budget of iodine is dominated by marine carbonates (). Continental deposits of iodine are thus ultimately of marine origin () and can inject iodine to the atmosphere as part of volcanic eruptions (, ) and lofted dust. The relative importance of atmospheric deposition and geogenic iodine as sources of iodate injected to the atmosphere remains to be established.

The correlation of iodate with calcium (fig. S6; see the Supplementary Materials for details) suggests that iodate reduction could be important to consider in geoengineering scenarios, which have proposed annual stratospheric injections of calcium carbonate as large as 5.6 Tg year−1 (). Iodate is a contaminant component of a variety of carbonate minerals (), i.e., 1 to 10 parts per million (ppm) (). Upper limits could lead to 3% year−1 increase in the current stratospheric iodine burden of 1.7 Gg (). Over a decade, iodate injection as part of geoengineering dust injections thus has the potential to significantly increase the iodine burden in the lower stratosphere, which would slow ozone layer recovery.

Our results have broader implications for iodine partitioning between the gas and particle phases and add field evidence of iodate reduction as a missing component of the geochemical iodine cycle. Multiyear records show that iodine has increased by a factor of 3 since 1950, responding to O3 anthropogenic pollution and thinning of polar sea ice (–). Atmospheric models are currently missing sources of iodic acid (HIO3), which is a highly condensable vapor that adds iodate in particles and—as an alternative pathway to higher iodine oxides—can nucleate and grow particles efficiently (). If atmospheric deposition of iodate is responsible for iodine accumulation on dust, it is currently unclear whether the released iodine we observe had accumulated before dust lofting on geological time scales or reflects the contemporary O3 response of the oceanic iodine source enhancement. The iodine-mediated feedback between dust and tropospheric O3 is likely to shift regional impacts relevant to air quality and human health in a changing climate. Predicting these changes in response to changing dust emissions on climate is challenging (, ) and deserves further attention. Investigation of the time scales of iodate cycling between the gas and particle phases, including deposition to arid regions and reemission as dust, is critical to quantifying the role of iodate and deserts as a missing piece in the global geochemical iodine cycle.

MATERIALS AND METHODS

The TORERO field campaign

The aircraft measurements were collected aboard the National Science Foundation (NSF)/National Center for Atmospheric Research (NCAR) High-performance Instrumented Airborne Platform for Environmental Research (HIAPER) GV aircraft during the Tropical Ocean tRoposphere Exchange of Reactive halogen species and Oxygenated volatile organic compounds (VOC) (TORERO) field campaign (). TORERO conducted 17 research flights over the Eastern Pacific Ocean out of Antofagasta, Chile and San Jose, Costa Rica in January and February 2012. IO was measured with a CU AMAX-DOAS instrument (, , ) (see the Supplementary Materials). Aerosol backscatter and depolarization profiles were measured by the HSRL (, ) (see the Supplementary Materials). Submicrometer particles were measured with an Ultra-High Sensitivity Aerosol Spectrometer (UHSAS) (), and supermicrometer particles were measured with a Cloud Droplet Probe (CDP) (). The up- and downwelling spectral actinic flux was measured by the HIAPER Airborne Radiation Package (HARP)—Actinic Flux () and used to compute photolysis frequencies. Organic iodine (particularly CH3I) was measured with the NCAR Trace Organic Gas Analyzer (TOGA) (, ).

Identification and specification of dust layers

To ensure a consistent identification and definition of the dust layers and their extent, three criteria were used: (i) The aerosol surface area, when summed from the UHSAS and CDP assuming spherical particles, was greater than 25 to prevent extensive splitting of layers if the surface area drops below this threshold for less than 30 s, the layer is treated as contiguous; (ii) IO mixing ratios in the layer either are significantly greater than measured above or below the layer or are greater and at least 0.2 pptv when averaged; (iii) water and potential temperature profiles were examined to exclude the boundary layer. Using this definition, a total of 27 layers, ranging in altitude between 0.9 and 6.4 km, were identified during the TORERO campaign. All identified layers are from flights that landed or took off from Antofagasta, Chile. For a further classification of the layers and the naming convention, see the Supplementary Materials.

Back-trajectory data products

Two back-trajectory data products were used: (i) a 72-hour mesoscale product in the Weather Research and Forecasting (WRF) model () and (ii) a 10-day Real-time Air Quality Modeling System (RAQMS) product. The WRF back-trajectories were initialized along the flight track for ~2000 points per flight and ran back for 72 hours at 3-hour resolution with an advective time step of 30 min (). The WRF simulations included a planetary boundary layer (PBL) tracer that was set to 1 within the model’s PBL and allowed to spread above the PBL by various model processes. Surface source regions for a given trajectory were defined as the locations where the PBL tracer increased or the trajectory passed through the boundary layer. Boundary layer contributions were collected for 1° × 1° bins for every back trajectory ending in an iodine enhancement layer back 72 hours weighted by the amount of tracer increase (weight of 1 when in the boundary layer). Dust source regions are determined by selecting source areas with soil moisture below 0.075 cm3 cm−3 at 1° × 1° resolution in the annual average based on the Soil Moisture Active Passive (SMAP) satellite (). Reverse Domain Filling () in RAQMS chemical and meteorological forecasts () provides information on 10 days of air mass history for ~500 altitude curtains per flight along the flight track.

Chemical box modeling

Iy,gas was inferred from IO radical measurements using a chemical box model developed at the University of Colorado Boulder (, ), with extended iodine chemistry () including photolysis frequencies measured with HARP—Actinic Flux () and chemical constraints from observations (see the Supplementary Materials for details). O3 loss for different chemical scenarios was modeled using the same box model (table S3).

Global modeling

CAM-chem (version 4) () was used to estimate ozone loss and the iodine (Iy) budget. Model cases examining the recycling of iodine to the gas phase on different atmospheric surfaces are the same as in (). The dust iodine source was represented as an extension of the iodine (IONO2, INO2, and HOI) recycling on sea-salt aerosol, although instead of just resulting in a change of Iy partitioning, the iodine recycling on dust surfaces produces an additional iodine source of 4% of the net heterogeneous recycling rate. The rate of this reaction was capped to saturate at 2 μm2 cm−3 and further limited by latitude such that IO all over the globe never exceeds the maximum concentration observed during TORERO within the SAAD region (see the Supplementary Materials for details).

Statistical analysis

Unless otherwise specified, when referring to fitted correlations in the text, the input data are the specified quantities arithmetically averaged over layer intercepts, and orthogonal distance regression is used. Data are used without additional weighting, if data appear on a graph showing a correlation they were used in the fit. No composite or nonlinear fits are presented in this manner; hence, references to “R” or “R2” refer to the Pearson product-moment correlation coefficient and the square thereof. Explicit hypothesis testing is not used.