Photochemical Modeling of Wintertime HO _x Sources and Sinks in a Subarctic Urban Location

Abstract

The hydroxyl radical (OH) is a key atmospheric oxidant that is important for regulating pollutant concentrations and particle formation and composition. The globally dominant formation mechanism of OH is ozone photolysis. However, the low-light conditions in the wintertime Arctic may limit OH formation through this pathway, and there is a lack of observations available to explain oxidant chemistry in polar locations. We use the Dynamically Simple Model of Atmospheric Chemical Complexity (DSMACC) 0D model to assess processes controlling the abundances of atmospheric oxidants in a subarctic city. To constrain the model, we use measurements made during the Alaskan Layered Pollution and Chemical Analysis (ALPACA) field campaign, which took place in Fairbanks, Alaska, USA, during the winter of 2022. HONO photolysis to form OH is the major chemical source of HO _x in the polluted, low-light environment of wintertime Fairbanks, and formaldehyde is an important precursor for HO₂. We find that the HO _x budget is distinct between a polluted, strongly stable inversion event and a clean, weakly stable period because of the different meteorology and pollution levels regulating peroxynitric acid (PNA, HO₂NO₂) cycling. Our results improve our understanding of the unique atmospheric pollution chemistry under cold and low-light conditions.

1. Introduction

Atmospheric oxidants are key players in chemical reactions that result in the transformation and loss of atmospheric pollutants and the formation of secondary pollutants. These oxidants include the hydroxyl radical (OH), ozone (O₃), and nitrate radical (NO₃). The reaction with OH is the dominant loss process for most reactive gases in the atmosphere. , As such, OH regulates the conversion of primary emissions to secondary species, influences particulate matter concentrations through the formation and growth of secondary aerosols, and on a large scale determines the lifetime of the greenhouse gas methane. ,

In polar regions, the cold temperatures, low solar insolation, and high solar zenith angle during winter limit the formation of OH, , resulting in prolonged lifetimes of reactive pollutants relative to typical low- to midlatitude conditions. Production of OH through the globally dominant formation mechanism in and is expected to be low in polar winter due to limited solar insolation at short wavelengths, suppressing the photolysis of O₃ in , and cold temperatures limiting water vapor concentrations (H₂O), suppressing . − In several winter environments, has been found to dominate OH production because there is more availability of light at the longer wavelengths needed for HONO photolysis than at the shorter wavelengths for ozone photolysis. , However, the dominance of HONO photolysis has not been seen in all wintertime studies of OH. In addition, polluted conditions in some wintertime locations affect the formation of OH via and the availability of the HO _x reservoir species peroxynitric acid (PNA, HO₂NO₂) in , although these results have not been found in all locations. ,, There is a lack of field measurements to corroborate the expected chemical regimes in Arctic locations, and modeling of wintertime OH has revealed uncertainties in chemical mechanisms such as the role of heterogeneous chemistry and missing reactivity. ,−

$${\mathrm{O}}_3+h\nu (\le 320\mathrm{nm})\to \mathrm{O}({}^1\mathrm{D})+{\mathrm{O}}_2$$ R1

$$\mathrm{O}({}^1\mathrm{D})+{\mathrm{H}}_2\mathrm{O}\to 2\mathrm{OH}$$ R2

$$\mathrm{HONO}+h\nu (400\mathrm{nm})\to \mathrm{OH}+\mathrm{NO}$$ R3

$${\mathrm{HO}}_2+\mathrm{NO}\to \mathrm{OH}+{\mathrm{NO}}_2$$ R4

$${\mathrm{NO}}_2+{\mathrm{HO}}_2\leftrightarrow {\mathrm{HO}}_2{\mathrm{NO}}_2$$ R5

The distribution of wintertime surface observations of either OH concentration or reactivity has a marked focus on northern-hemisphere, midlatitude locations (Figure ). Aircraft flight campaigns such as TOPSE or ATom , have measured wintertime concentrations aloft rather than at the surface; a comprehensive comparison of OH measurements made during flight campaigns is given in Miller and Brune. Several recent surface-based field studies have focused on Beijing and the North China Plain due to their intense wintertime haze events. ,, There are also winter measurements in urban areas such as New York City, Tokyo, and Birmingham, UK, ,,,− as well as rural areas. ,− Some studies have demonstrated differences between winter and summer oxidant concentrations and VOC lifetimes in given locations. ,,,, However, none of these studies capture the extreme cold and dark conditions experienced in polar regions during winter.

1.

Latitudes of field studies that included measurements of surface level atmospheric OH concentration (blue circles), reactivity (red crosses), or both (brown squares). The name of the field campaign or the measurement location is listed with the reference for the OH measurements. Only campaigns that included surface wintertime (DJF) measurements are included. We found no studies in the southern hemisphere of wintertime (JJA) measurements and therefore did not include this region in the x-axis. Only campaigns that measured reactivity were included, and campaigns that calculated reactivity based on measured VOC concentrations were not. The RONOCO campaign, marked with *, intended to measure OH but concentrations were lower than the limit of detection and Walker et al. gives an upper limit to the OH concentration for this campaign.

The Alaskan Layered Pollution and Chemical Analysis (ALPACA) 2022 field project took place from January to March 2022 in Fairbanks, Alaska, USA, with the goal of better understanding atmospheric chemistry during high-latitude winter in urban environments. Fairbanks is located at 64.8 °N and experiences poor air quality each winter. Strong and persistent surface-based temperature inversions trap local emissions and often lead to PM_2.5 concentrations that exceed the US Environmental Protection Agency (EPA) ambient air quality standards. , There are uncertainties surrounding the chemistry that occurs during these pollution events. Specific open research questions include: which oxidants are present under cold and dark conditions, what are their roles in SOA formation and sulfur oxidation, and how does reactive nitrogen chemistry affect oxidant concentrations? Some studies based on data from the ALPACA project have examined the aspects of oxidant chemistry in Fairbanks, e.g., Farrell et al., but as of yet there have been no modeling studies of the chemistry controlling OH during the campaign.

Here, we examine oxidant chemistry in Fairbanks during the winter, with a primary focus on the OH and HO₂ (HO _x ) budgets using a combination of chemical modeling and field measurements from ALPACA 2022. We estimate concentrations of HO _x during ALPACA using an established box modeling framework. We use the Dynamically Simple Model of Atmospheric Chemical Complexity (DSMACC), a zero-dimensional model, , constrained with measurements made during ALPACA 2022 to simulate near-surface photochemistry during the campaign period. The goal of this work is to identify key atmospheric processes that affect the oxidation capacity and secondary air pollution formation in urban polar winter conditions.

2. Materials and Methods

2.1. Model Setup

To simulate near-surface photochemistry in Fairbanks, we used the Dynamically Simple Model of Atmospheric Chemical Complexity (DSMACC). , We used the Master Chemical Mechanism (MCM) v3.3.1, , available at https://www.mcm.york.ac.uk (last access: Nov 17, 2025) processed with Kinetic PreProcessor (KPP) v2.2.3. Clear-sky photolysis rates were calculated using Tropospheric Ultraviolet and Visible (TUV) Radiation Model v4.2; the measured J _O¹D and J_NO2 from the ALPACA campaign were used to scale other clear-sky photolysis rates to account for cloud cover.

We constrained the model by using measurements made during the ALPACA campaign. These measurements included available VOCs (listed in Table ) measured by proton transfer reaction time-of-flight-mass spectrometry (PTR-ToF-MS); , formaldehyde measured by COFFEE (compact formaldehyde fluorescence experiment) and Aeris (mid-infrared absorption) instruments; , NO, NO₂, SO₂, CO, and O₃ by Thermo Scientific gas analyzers; and HONO concentration measured aloft (from 12 to 17 m) by long-path differential optical absorption spectroscopy (LP-DOAS). For each m/z detected by the PTR-ToF-MS, major VOCs contributing to the signal at the corresponding ionic formula to organic compounds were assigned based on previous references for VOC emission profiles associated with sources identified in Fairbanks in wintertime, thus reducing to the subset of representative compounds listed in Table . All concentrations except HONO were measured at the same ALPACA supersite in downtown Fairbanks, while the HONO measurements were made nearby. We resampled observations at a frequency of 20 min to correspond to the HONO measurements, which had the coarsest temporal resolution. Model simulations were conducted for periods when valid measurement data of all inputs were available, giving a total of 593 modeled data points across the 40 day campaign. Figure shows a time series of a subset of the input data used to constrain the box model. The gaps in the time series are due to one or more missing inputs for each 20 min of resampling interval. The resulting model output is consistent with this 20 min frequency, including gaps in the time series (i.e., the model does not interpolate between data points). The distributions of CO mixing ratios measured during the entire ALPACA campaign period and for the subset time series of data presented in Figure agree well (see Supporting Information Figure S1). This implies that our modeling results are representative of the conditions sampled throughout the ALPACA campaign.

1. Average Concentrations of VOCs Included as Constraining Species.

		percentile
gas	mean (ppb)	10th (ppb)	90th (ppb)
constrained in all simulations
formaldehyde	3.972	0.983	9.08
methanol	5.394	1.104	10.81
acetaldehyde	1.721	0.627	3.09
formic acid	0.915	0.580	1.30
ethanol	3.827	1.025	6.51
acetone	1.249	0.548	2.10
acetic acid	1.973	0.445	3.70
MVK + MACR	0.171	0.030	0.37
butanone	0.179	0.098	0.28
benzene	0.952	0.167	2.17
2,3-butanedione	0.180	0.050	0.32
toluene	2.255	0.239	5.42
phenol	0.086	0.007	0.18
acetic anhydride	0.027	0.009	0.05
styrene	0.074	0.007	0.17
2,5-hexanedione	0.016	0.003	0.03
benzoic acid	0.034	0.003	0.08
constrained in sensitivity analyses
isoprene	0.414	0.061	0.87
ethane	2.591	1.081	4.54
ethene	15.383	3.234	31.73
propane	0.740	0.318	1.36
propene	5.164	1.043	10.67
n-butane	0.232	0.076	0.46
i-butane	0.102	0.031	0.22

^aMeasured by COFFEE and Aeris, not by PTR-ToF-MS.

^bPTR-ToF-MS signal reported as MVK + MACR, represented as only MVK in the model.

^cThe PTR-ToF-MS ion was not conclusively identified.

^dEstimated concentration.

2.

Time series of a subset of inputs used in the box model to constrain the simulations. Surface area (orange) and temperature (black) in the top row, ozone (pink) and J _{O ¹D} (black) in the middle row, and NO₂ (cyan) and J_NO2 (black) on the bottom row. The shading denotes the two time periods selected for case study analysis. Period 1 (Jan 29 through Feb 02) in gray shading is a polluted period, during which NO₂ and particle pollution were high but ozone was depleted. Period 2 (Feb 05 through Feb 09) in yellow is a clean period, when NO₂ and particle pollution are low but ozone reaches ambient concentrations upward of 35–40 ppb. Sunlight increases throughout the campaign period, as represented by the increasing and broadening J_O¹D and J_NO2 .

As shown in Figure , there are some significant gaps in the temporal coverage of the model inputs. We therefore chose two case study periods during which the temporal frequency of model inputs was high and conducted a detailed comparison of these two periods to demonstrate the variability of the HO _x chemistry throughout the campaign. Period 1 (Jan 29 through Feb 02, gray-highlighted period in Figure ) was in the middle of a strong surface-based inversion event, which resulted in high PM_2.5 and NO _x concentrations, and we refer to this period as the polluted period. , We considered Period 2 (Feb 05 through Feb 09, yellow-highlighted period in Figure ) a clean event. During this period, there was more advection and mixing from aloft and only weakly stable conditions. The two case study periods occurred during the first half of the ALPACA campaign, coinciding with low sunlight levels and cold temperatures. However, the average temperature during Period 2 (clean period) was about 10 K warmer than that during Period 1 (polluted period), 245 K compared to 254 K.

The DSMACC model does not explicitly simulate nonchemical loss of intermediate species via physical processes such as deposition, mixing, and advection. To prevent an unrealistic accumulation of the species not constrained by measurements, we add a first-order loss term (k _phys) for all species with a rate coefficient of 1.16 × 10^–5 s ^–1 equivalent to a lifetime of approximately 24 h. This is based on the work of Stone et al., which showed that HO _x concentrations were not significantly impacted by changes to deposition lifetime in the model. We address this assumption by testing the sensitivity of the model to a range of k _phys values. As the MCM only comprises gas-phase chemical reactions, we introduce heterogeneous loss for OH and HO₂ onto aerosol surfaces as first-order loss terms (eq ) where k _het is the heterogeneous loss first-order rate coefficient, SA is the effective surface area per unit volume (cm²/cm³), which we constrained in the model with measurements made by a scanning mobility particle sizer (SMPS), and γ _i is the unitless uptake coefficient of gas i. The mean molecular speed for each gas c _i is shown in eq , where R = 8.31451 J·mol^–1·K^–1 is the ideal gas constant, T is the temperature in Kelvin, and M _w is the molar mass of the gas. We include γ_HO2 = 0.1 and γ_OH = 1.0 as the default simulation values and test the sensitivity of the model to a range of uptake coefficient values.

$$k_{\mathrm{het}}=\frac{c_i\mathit{SA}{\gamma }_i}{4}$$ 1

$$c=\sqrt{\frac{8\mathit{RT}}{\pi M_{\mathrm{w}}}}$$ 2

The relative humidity in the base simulation was set to 72% as an average of the available measurements from Fairbanks International Airport during the ALPACA project. The water vapor volume mixing ratio was calculated from the average relative humidity, observed pressure collected at the Fairbanks International Airport, and the observed temperature from the downtown ALPACA supersite. The pressure measurements were made on an hourly basis, so we resampled them to 20 min frequency using a linear interpolation to calculate water vapor volume mixing ratio. The mixing ratio of CH₄ was set to a constant value of 2019 ppb based on the average of available in situ and flask observations from January through February made at the Barrow, Alaska, and Cold Bay, Alaska, NOAA Carbon Cycle Cooperative Global Air Sampling Network sites. We set H₂ to a constant mixing ratio of 516 ppb, as the average of flask measurements made between January and February 2022 from the Barrow and Cold Bay sites. Broadband albedo was set to 0.7, based on a mean of albedo values calculated from observed J _NO2 radiation fluxes during the campaign (mean 0.66, minimum 0.54, and maximum 0.99).

2.2. Sensitivity Analysis

We conducted several sensitivity experiments to better understand the factors influencing our model. We tested the sensitivity of the model-simulated HO _x to loss via physical processes (k _phys), HONO concentration, relative humidity, uptake of HO₂ and OH to particles (γ_HO2 and γ_OH), and constrained VOC inputs. For all sensitivity analyses, we evaluated the change in midday (10 am to 2 pm) local Alaska Standard Time (AKST) OH and HO₂ concentrations.

We adjusted k _phys for all species in two sensitivity experiments: decreasing k _phys so that the lifetime with respect to physical loss would be approximately 5 days and increasing k _phys so the lifetime would be approximately 1 h. We conducted a third sensitivity experiment in which we adjusted only the physical lifetime of PNA to 1 h, k _phys,PNA = 2.78 × 10^–4 s ^–1, to isolate the impact of PNA, and investigate how its potential accumulation or dispersal may affect the HO _x budget.

We adjusted HONO concentration in two sensitivity tests by ±20% to evaluate the impact of uncertainty in the observed HONO. All measurements used in our modeling aside from HONO are made at the surface. The remote sensing measurements sample HONO concentrations from 12 to 17 m height above the surface. We expect that the degree to which the ambient HONO concentration in Fairbanks is well-mixed and representative between the surface and aloft will be sensitive to vertical stability and may differ between the polluted (stable) and cleaner (weakly stable) periods.

We conducted three sensitivity tests with RH set to 30, 60, and 90% in the calculation of the water vapor mixing ratio. These values cover the range of RHs measured at the Fairbanks International Airport. Relative humidity and pressure were used to calculate the water vapor mixing ratio. The pressure measurements we used were interpolated from an hourly frequency to 20 min to align with other ALPACA measurements. These three RH sensitivity tests allowed us to evaluate both the impact of interpolating the coarser resolution data and the impact of using a fixed average RH on model output, as the change in water vapor mixing ratio from changing RH was greater than the change introduced by adjusting pressure.

We tested the impact of removing OH uptake (γ_OH = 0) and decreasing γ_HO2 from 0.1 in the base simulation to 0.01 and 0.001. Recent laboratory studies have found that OH reactive uptake depends on a number of factors including particle composition, temperature, and relative humidity. − Reactive uptake and heterogeneous oxidation by OH are considered important drivers of organic aerosol chemical aging. ,, Because of the wide range of γ _OH values in the literature, , we tested the sensitivity of our model using extremes of this range. There is also a wide range of γ_HO2 values in the literature from lab, modeling, and field studies, which show dependence on factors such as particle composition, pH, ambient temperature, and NO _x concentration. ,− A typical value used in models is γ_HO2 = 0.2 based on observations made and comparison with a model during the 1997 SONEX field campaign. , However, later field measurements have found that γ_HO2 can vary widely even over a short period in one location. − Zhou et al. recommended using different uptake values in models to be representative of the area of interest, but there is limited analysis of wintertime HO₂ uptake in the Arctic. To account for the lack of available literature values, we tested γ_HO2 over 3 orders of magnitude.

In our default model setup, we used a limited number of VOCs to constrain the model based on available measurements. First, we assumed that the C ₅H₉ ion measured by PTR-ToF-MS was not isoprene based on the recent literature, suggesting that fragments of aldehydes and cycloalkanes from anthropogenic sources including fossil fuels and cooking can interfere with the detection of C₅H₉ by PTR-MS. , In Fairbanks, enhancement in the C₅H₉ ion is associated with home heating oil and traffic. There is also likely a negligible contribution from biogenic isoprene in this winter urban environment. We therefore excluded this species as a constraint in the base model setup. As the C₅H₉ fragment in the ALPACA measurements has not been conclusively attributed to isoprene or another VOC, we performed a second simulation with an identical setup and inputs to our base simulation but with the model constrained to C₅H₉ concentration under the assumption that it was isoprene. Because isoprene has higher reactivity with OH than other species that produce the C₅H₉ fragment such as octanal and nonanal, , this constrained simulation (called isoprene constrained, IC) and the base simulation represent extremes of the potential impact of the C₅H₉ -based constraint on the HO _x budget and concentrations.

In addition, C ₂–C ₄ alkanes and alkenes were not measured by PTR-ToF-MS and therefore we did not include these species in our base model simulation. We conducted a sensitivity test with estimated ethane, ethene, propane, propene, i-butane, and n-butane concentrations (Table ) added to the model as constraining species. We estimated the C ₂–C ₄ alkane and alkene concentrations using emission ratios (ERs) from the major sources of VOC emissions in Fairbanks. Our calculation of VOC concentrations using these ERs is intended to allow a simple sensitivity of model HO _x to potential sources of additional VOCs in Fairbanks. We focused on the primary emission sources of biomass burning (i.e., wood combustion), heating oil, and traffic, which were identified by PMF analysis as being important factors in Fairbanks by Ketcherside et al. We use ERs from Permar et al. for the long-lived biomass burning factor. The published ERs are in ppb VOC per ppm CO, which we converted to ppb VOC per ppb formic acid, as about 90% of the measured formic acid in Fairbanks is from biomass burning. For traffic emissions, we used measured emission factors from Gordon et al., which are reported in mg/kg of fuel. We convert these emission factors to ERs of ppb VOC per ppb NO _x , as about 55% of NO _x in Fairbanks was from vehicles. Emissions of VOCs from vehicles depends on the engine type, fuel type, driving conditions, and vehicle maintenance − and measured emissions has varied between locations and over time. − In general, C ₂–C ₄ alkanes and alkenes are a small contribution of vehicle exhaust, although some species are a larger fraction of fuel evaporation, which we do not consider in our calculations. ,, Finally, for fuel oil, we used measurements from a residential oil burner study reported in mg/kg fuel. We converted these emission factors to ERs of ppb VOC per ppb benzene, as heating oil was a major source (approximately 35%) of benzene during the ALPACA project. For each source, we weight the representative species concentration C _rep (i.e., formic acid, NO _x , and benzene) by the relative contribution of the source they represent as determined by Ketcherside et al. (PMFfraction) to calculate the species concentrations C _i as shown in eq .

$$C_i=C_{\mathrm{rep}}\times \mathrm{PMFfraction}\times E{R}_i$$ 3

Adding the new VOCs to our constrained species required that the mechanism be updated for all MCM reactions that included these species and their intermediates. We tested the new mechanism with the base set of constrained concentrations, excluding the six new VOCs and isoprene, to isolate the impact of adding further reactions to the mechanism. There was an average of −0.03% difference in both OH and HO₂ concentrations between the updated mechanism simulation and the base simulation, both run with the same constraining species. Thus, just adding reactions with the new VOCs without constraining these species had limited impact on the model.

3. Results and Discussion

3.1. HO _x Concentrations in Fairbanks

Concentrations of HO _x and PNA throughout the ALPACA campaign calculated by DSMACC are shown in Figure . Concentrations of OH and HO₂ are low in January and start to increase during early- to mid-February as sunlight increases. There were more than four additional hours of sunlight by the end of the ALPACA campaign compared to the beginning. OH has a clear dependence on the time of day, with a peak in average diel concentration of 3.7 × 10⁵ cm^–3 at 1 pm and close to zero between 5 pm and 8 am. The HO₂ concentration is higher through much of the day compared to OH but shows little variation with sunlight. Peak average diel HO₂ concentration is 1.4 × 10⁶ cm^–3 at 3 pm, while the concentration at noon is half of this. During some periods of the campaign, such as from Feb 14 to Feb 24, HO₂ is maximum during the day, in line with OH. During other periods, such as between Feb 04 and Feb 09, there are large spikes of HO₂ during the night. This results in a large range in diel HO₂ concentration, with the exception of 8 am local time, when day-to-day variability in HO₂ is minimal. This dip in HO₂ variability is due to depletion overnight prior to the beginning of photolysis-driven production and will be explored further in Section . The total HO _x concentration follows a similar pattern as HO₂, but with a small increase in mean concentration at 1 pm relative to HO₂ because of the peak in OH concentration at this time. The concentration of PNA peaks in the early evening and decreases to a minimum in the morning. This diel trend is quite clear during the polluted period from Jan 29 to Feb 03 in the time series, but the diel cycle of PNA concentrations becomes more variable later in the time series. From Feb 05 onward, PNA concentration tracks with HO₂, potentially contributing to the elevated concentrations of HO₂ at night. This relationship will be explored more in the budget analysis in Section .

3.

Time series of OH and HO₂ (top) and HO _x and peroxynitric acid (PNA) (middle) during the ALPACA campaign. Mean diels are plotted along the bottom, with shading to show the 10th and 90th percentile of concentration each hour. The shaded regions in the upper panels denote the time range for the case study periods: Period 1 in gray and Period 2 in yellow.

Our modeled concentrations of HO _x in Fairbanks are lower than measurements made in most urban locations in winter. Measured wintertime concentrations of OH were on the order of 2 × 10⁶ cm^–3 or greater in Beijing, ,, New York City, Tokyo, Boulder, CO, and Birmingham, UK. The two most northerly campaigns shown in Figure report OH concentrations of a similar magnitude as calculated in our model. During the RONOCO campaign in the UK, nocturnal wintertime OH concentrations were below the instrument detection limit, and the estimated maximum OH concentration at night was 6.4 × 10⁵ cm^–3. During the COBRA campaign in rural Canada, peak noon concentrations were 1.16 × 10⁶ cm^–3 but were below the detection limit (6.4 × 10⁵ cm^–3) for much of the rest of the day. Wintertime HO₂ concentrations ranging from 0.4 × 10⁸ to 4.5 × 10⁸ cm^–3 were made in Beijing, Tokyo, New York City, and Birmingham, ,,,,, which are 1–2 orders of magnitude greater than those calculated in our model. HO₂ concentration in the COBRA campaign was also greater than that calculated here, peaking at 1.42 × 10⁸ cm^–3 at noon, closer to the other urban measurements. We explore the drivers of the estimated HO _x concentrations simulated in Fairbanks in our model budget analysis.

3.1.1. Comparison of Clean and Polluted Periods

We compared the OH, HO₂, and PNA concentrations during two different periods of the campaign to better understand the influence of pollution abundance on the HO _x chemistry in Fairbanks. The diels in Figure demonstrate that concentrations were distinctly different between the two periods. During the polluted period, the level of NO₂ is greater than 40 ppb for the entire day. The build-up of NO₂ coincides with near-zero O₃ concentration as ambient O₃ is titrated with NO. This chemistry during the polluted event is driven by stable atmospheric conditions that trap local NO _x emissions but prevent mixing with O₃-rich air from outside Fairbanks. , In comparison, during the clean event, weakly stable meteorology allows for mixing- and transport-driven dispersion of local emissions, reducing the NO _x concentrations. Thus, while we describe the impact of different chemical regimes on HO _x , these regimes are influenced by varying meteorological conditions across the campaign period.

4.

Diel concentrations of selected species for the clean (columns 2 and 4) and polluted (columns 1 and 3) periods. Concentrations for constraining inputs are shown in the left two columns: ozone (pink, top row), NO₂ (cyan, middle row), and HONO (purple, bottom row). Model outputs are shown in the right two columns: OH (blue, top row), HO₂ (brown, middle row), and PNA (green, bottom row). Species are plotted on the same y-axis scale between the polluted and clean periods except for HO₂ because the concentration during the polluted period is an order of magnitude smaller than during the clean period. Shading for all species represents the 10th and 90th percentiles.

HO _x concentrations are lower on average during the polluted period compared to the clean period, and HO₂ concentration in the clean period follows a similar diel cycle as OH with a peak in concentration around midday. During the clean period, O₃ concentration is high, sometimes reaching 35–40 ppb, and is consistent with concentrations in ambient air masses outside the Fairbanks urban area. The NO₂ concentration is much lower during the clean period than during the polluted period. The HO₂ concentration does not show a strong daytime peak or dependence on the time of day in the clean period. In both clean and polluted periods, the HO₂ concentration is at a minimum at 8 am, which was reflected in the small concentration variability in the campaign average diel in Figure . The OH diel profiles are broadly similar between the two periods, despite differences in HO₂. Studies in Beijing have also reported a similar OH behavior during haze versus nonhaze events. , However, during these studies, HO₂ concentrations were also similar during haze and nonhaze events, , indicating that there may be different pollution sources and chemistry occurring in Beijing haze events compared to Fairbanks pollution events. The PNA diels are also broadly similar between the two periods despite the differences in NO₂ and HO₂ concentrations. The stark differences between HO _x and PNA concentrations between the clean and polluted periods hint at different chemical regimes, which will be explored further in budget analysis.

3.2. HO _x Budget

The major chemical production and loss reactions in the box model for the campaign average are shown in the rate of production and loss analyses (ROPAs) in Figure . Production of OH in Fairbanks is dominated by HONO photolysis, which peaks at around midday. The reaction of HO₂ with NO is the second-largest source of OH, and this reaction continues to produce OH after sunset. The loss of OH is dominated by the reaction with NO₂ to form HNO₃. There is also some recycling of NO and OH to reform HONO, although loss by this mechanism does not balance with the production of OH from HONO. Reactions with acetaldehyde and CO are also sinks of OH during the day. Acetaldehyde during the ALPACA campaign is mainly produced from wood burning and heating oil combustion.

5.

Rate of production and loss analysis (ROPA) with the top five production (positive) and loss (negative) reactions in the model for OH (top), HO₂ (middle), and HO _x (bottom).

In comparison to OH, there are a variety of HO₂ sources during the day. The sum of HCHO reactions contributes the most to HO₂ production when there is available sunlight, although OH + CO and CH₃O + O₂ are also important formation mechanisms. Formaldehyde comes from a mixture of sources in Fairbanks, including local wood burning, heating oil combustion, traffic emissions, and secondary formation during aging of polluted air masses. The loss of HO₂ is dominated by the reaction with NO to form OH. The formation of PNA is the second-largest loss term for HO₂. At night, HO₂ formation from PNA is greater than its loss, meaning that there is a net production of HO₂ from PNA. A similar behavior of PNA was reported during the COBRA campaign in rural Canada. The net production of HO₂ via PNA then contributes to OH production at night through the reaction with NO. However, if the residence time of PNA is short, then, this nocturnal production pathway breaks down.

When the lifetime of PNA with respect to physical loss was set to approximately 1 h in our sensitivity analysis, there was only minor impact on midday HO _x concentrations, but the impact on nighttime HO₂ concentration and the budget was large (Figure ). With the PNA residence time reduced, the nocturnal chemistry is suppressed and HO₂ concentration goes to zero. The OH budget is also impacted, as there is no OH produced from HO₂ at night, but the OH concentrations are similar between the two simulations. The physical residence time is largely determined by horizontal advection and mixing with air above the polluted surface layer. Residence time is expected to vary between stable, polluted periods and less stable, cleaner periods. While the sustained production of HO₂ via PNA was also seen in rural Canada in the winter, the lack of PNA or HO _x measurements to constrain our model makes it difficult to verify how important PNA cycling is in Fairbanks. Adjusting the k _phys of PNA implies that the physical residence time of PNA is the key determinant of its role in the nighttime HO _x budget.

6.

ROPAs (left y-axis) of the largest production (positive) and loss (negative) reactions in the chemical mechanism for OH (top) and HO₂ (bottom) and concentration (right y-axis) of OH and HO₂ for the base simulation (solid line) and the simulation with k _phys,PNA set to a residence time of approximately 1 h (dashed line).

The third row of Figure shows the net total HO _x budget. The reactions that recycle between OH and HO₂, such as HO₂ + NO → OH + NO₂, do not contribute to net HO _x loss or production. HONO photolysis is the dominant net source of HO _x , contributing on average 64% to daytime HO _x production. Photolysis of HONO is an important HO _x source in other winter locations, for example, Beijing, , Boulder, CO, and New York City. In our model, chemical HONO production occurs through two reactions: NO + OH and the heterogeneous reaction of NO₂ with particles. The contribution of the NO₂ heterogeneous reaction to HONO concentration is small in our simulations, although this may be due to the simplification of the reaction with a constant uptake coefficient. The reaction between NO and OH contributes more to HONO but is often less than the amount of HONO lost via photolysis, as demonstrated in Figure . This implies that other processes in addition to chemical production are important for controlling the abundance of HONO in Fairbanks, such as direct local emissions, transport of existing HONO from regional sources, or additional heterogeneous production processes.

Alkene ozonolysis was reported in previous campaigns such as PUMA and BEST-ONE as being an important HO _x production pathway, , but this pathway was not significant in our model. The dominant sink of HO _x in our model is OH + NO₂ → HNO₃, accounting for 33% of total HO _x loss on average and up to 50% at midday. Formation of HNO₃ was the largest sink of HO _x during the BEST-ONE campaign, near Beijing and in New York City during the winter PMTACS-NY campaign. Overall, our modeled HO _x budget is dominated by features from the OH budget, while HO₂ formation and loss pathways are small contributions.

3.2.1. Sensitivity of the Model Budget to Constrained Inputs

Here, we present results of the sensitivity experiments to examine the impacts of key assumptions made in our box modeling framework (Table ).

2. Sensitivity of Midday HO _x Concentration to Model Perturbations.

base simulation input	modification	ΔOH (%)	ΔHO2 (%)
k phys = 1.16 × 10–5s–1 (τphys ≈ 1 day)	k phys = 2.78 × 10–4 s –1	–3.00	–10.70
	(τphys ≈ 1hr)
	k phys = 2.31 × 10–6 s –1	1.39	5.82
	(τphys ≈ 5day)
	k phys,PNA = 2.78 × 10–4 s –1	–3.12	–7.19
	(τphys,PNA ≈ 1hr)
[HONO] from DOAS	1.2 × [HONO]	14.02	7.87
	0.8 × [HONO]	–14.10	–7.94
RH = 72%	RH = 30%	–0.07	–0.03
	RH = 60%	–0.02	–0.01
	RH = 90%	0.03	0.01
γOH = 1.0	γOH = 0.0	0.50	0.30
γHO2 = 0.1	γHO2 = 0.01	0.11	0.24
	γHO2 = 0.001	0.12	0.26
PTR-ToF-MS-identified VOCs	isoprene	–0.14	9.22
	C 2–C 4 alkanes/alkenes	–11.19	58.21

One uncertainty to our analysis of the HO _x budget during ALPACA is the limited available VOC measurements. When we constrain with additional estimated C ₂–C ₄ alkanes and alkenes, midday OH concentration decreases (−11% from the base model) and midday HO₂ concentration increases (+58%). The decrease in OH concentration is due to additional reactivity with propene and ethene, which become the third and fourth most important OH loss mechanisms, respectively (Figure ). The addition of the C ₂–C ₄ alkanes and alkenes also increased OH production from HO₂+NO, as HO₂ concentration is increased, but no other OH production reactions were significantly affected. In comparison, the addition of the C ₂–C ₄ alkanes and alkenes enhanced HO₂ production during the day but did not contribute to additional loss except through the HO₂+NO reaction. The decompositions of the 1-hydroxypropane-2-yloxy radical and the 2-hydroxyethoxy radical, formed from the oxidation of propene and ethene by OH, are the third and fourth largest contributors to HO₂ production with the additional VOCs, respectively.

7.

ROPAs (left y-axis) of the largest production (positive) and loss (negative) reactions in the chemical mechanism for OH (top) and HO₂ (bottom) and concentration (right y-axis) of OH and HO₂ for the base simulation (solid line) and the simulation with the updated VOC mechanism and constrained to additional VOC concentrations (add. VOC) (dashed line).

Previous studies using HO _x box modeling have argued that missing VOC measurements may help explain model-observation discrepancies, for example, by explaining missing OH reactivity. , However, some analyses in polluted winter conditions have found that short-chain alkane and alkene species are a small contribution to OH reactivity. , Our sensitivity analysis suggests that ethene and propene contribute to OH loss in Fairbanks, but short-chain alkanes do not. Missing alkene ozonolysis has also been proposed to explain OH underestimates compared to observations, , but we find that OH production is not significantly impacted by additional reactions. This is likely due to the low ozone concentrations during the ALPACA period. Overall, we find that our model is sensitive to available VOC measurements that are used to constrain the simulations, specifically, short-chain alkenes.

Adding isoprene to the model constraints had a mixed effect on HO _x concentrations at midday. OH changed by −0.14% on average compared to the base simulation and HO₂ increased by 9.2% (Table ). This increase is due to additional HO₂ production reactions: there are 464 reactions that contribute to the ‘Other’ category for HO₂ production in the IC simulation compared to 216 in the base simulation. The largest contributor to the ‘Other’ category in the IC simulation is the decomposition of the isoprene derivative ISOPBO (1-methyl-1-vinyl-2-hydroxyethoxy radical, CH₂(OH)­CO­(CH₃)­CHCH₂). This reaction did not contribute to HO₂ production in the base simulation. These additional HO₂ production pathways increase total HO₂ concentration, which then increases OH concentration by reaction with NO. The increased production of OH is offset by the additional OH lost through reaction with isoprene and its derivatives, resulting in the small, negative percent difference in midday concentration.

The adjustments to physical loss rate had a small mean impact on the midday HO _x concentrations. As expected, a faster k _phys reduced HO _x concentrations and a slower k _phys increased HO _x concentrations. During the day, when production of HO _x is high, changes in k _phys have a small relative impact on ambient concentrations, and hourly mean percent difference for both OH and HO₂ concentrations is small. The impact on HO₂ is greater than that on OH. This is in part because there are more species that contribute to HO₂ production than OH production, which is driven by constrained HONO concentration. A change in k _phys affects the concentrations of HO₂ reactants such as PNA, CO, and many of the species in the “Other” category, resulting in a large cumulative impact.

Model-simulated midday HO _x concentrations display high sensitivity to uncertainty in the input HONO concentration because of the importance of HONO in the OH budget. A 20% increase in HONO concentration (1.2 × [HONO]) results in an increase in midday OH, which is slightly sublinear (a 14.0% increase in OH concentration is produced by a 20% increase in HONO input). This behavior results from a 7.9% increase in HO₂ from cycling between OH and HO₂. A 20% decrease in HONO input (0.8 × [HONO]) also results in an 14.1% decrease in OH concentration and 7.9% decrease in HO₂. There is uncertainty if the HONO concentration aloft is representative of the surface conditions during weakly stable periods. These sensitivity experiments highlight the need for further study of HONO in the Arctic with colocated measurements.

Of the parameters tested, relative humidity had the smallest impact on HO _x concentrations. Varying relative humidity from 72% in the base case to 30, 60, or 90% resulted in less than ±0.1% difference in total HO _x concentrations on average. This result emphasizes the minor role of as a source of OH in Fairbanks winters (Figure ). Because of the minimal importance of , there is little impact of using a fixed average RH value and calculating the water vapor mixing ratio from interpolated hourly pressure measurements on our model results.

The difference in HO₂ concentration to changes in γ _{HO 2} was less than 3% between the smallest and largest uptake values, which implies that HO₂ loss in Fairbanks during winter is not sensitive to particle uptake. The uptake reaction in the model is included in the ‘Other’ category, which in Figure is a small contributor to total HO₂ loss when γ_HO2 = 0.1. Previous modeling studies of springtime subarctic conditions have found that HO₂ uptake to particles accounts for up to 30% of the HO _x loss. , However, the lower NO _x concentrations in these studies compared to the NO _x concentration measured during ALPACA mean that other loss reactions such as uptake become more competitive. In addition, γ_HO2 depends on several factors ,, and uptake values representative of the spring or summer may not be ideal for Arctic winter. Concentrations of OH and HO₂ in our analysis were also not sensitive to γ_OH. OH concentrations changed by less than 1% on average between simulations with γ_OH = 1 and γ_OH = 0, indicating that OH in our model is not sensitive to reactive uptake.

3.2.2. Budget in the Clean and Polluted Periods

While Figure shows the diel cycle of the major chemical reactions contributing to HO _x production and loss in Fairbanks, Figures and demonstrate that there are large differences in the OH and HO₂ budgets between the clean and polluted periods. There is enhanced OH production in the polluted period relative to the clean period and the campaign average in Figure . This increase is primarily from enhanced HONO photolysis despite the low-light conditions. There is also an increase in recycling from HO₂ via a reaction with excess NO. A few other reactions have appreciable contributions to OH production during either period. There is a slight increase in the level of OH production at night in the clean period relative to that in the polluted period. The OH production to loss ratio (P/L) for Period 1 is 1.01, meaning that the enhanced production during the polluted event is balanced by an enhanced loss. Recycling of OH back into HONO is the dominant loss mechanism in the polluted period, followed by the formation of HNO₃ by reaction with NO₂. During Period 2, there is excess production relative to loss of OH (P/L = 1.22). The formation of HNO₃ is the dominant loss mechanism, while the reaction with NO to reform HONO is minimal because there is less available NO during the clean period. In comparison, in the polluted period, NO _x emissions are trapped in the inversion, and there is minimal mixing with clean air from outside Fairbanks. Ozone is depleted through titration with NO to form NO₂; excess NO is therefore available to react with OH among other species.

8.

ROPAs of the largest production (positive) and loss (negative) reactions in the chemical mechanism for OH, comparing two time periods during the campaign. Jan 29 to Feb 02 is a polluted period during a strong surface-based inversion, while Feb 05 to Feb 09 was clean. We only explicitly show reactions that are significant contributors to either loss or production. All other reactions are included in ‘Other’. The number of samples in each hour bin is included along the top y-axis.

9.

ROPAs of the largest production (positive) and loss (negative) reactions in the chemical mechanism for HO₂, comparing two time periods during the campaign. Jan 29 to Feb 02 is a polluted period during a strong surface-based inversion, while Feb 05 to Feb 09 was clean. We only explicitly show reactions that are significant contributors to either loss or production. All other reactions are included in ‘Other’. The number of samples in each hour bin is included along the top y-axis.

There is enhanced daytime HO₂ production in the polluted period relative to the clean period (Figure ). However, at night, there is a distinct increase in HO₂ production in the clean period compared to the polluted period. This production is driven by decomposition of PNA. The enhanced HO₂ production at night does not, however, outweigh the enhanced loss during Period 2 (P/L = 0.91). The nocturnal production decreases throughout the evening, reaching a minimum shortly before sunrise. In the early morning, loss outweighs production in the clean period, resulting in the local minimum in HO₂ concentration in Figure . In the polluted period, HO₂ production and loss are balanced (P/L = 1.01). Cycling between PNA and HO₂ + NO₂ is not a large part of the budget during Period 1. Production of HO₂ during the polluted period is dominated by production from HCHO, methoxy radical (CH₃O), and smaller individual contributions from reactions lumped into ‘Other’. The lack of production via PNA in the polluted period at night results in near-zero HO₂ concentration for much of the evening and early morning. The differences in PNA chemistry between the two periods may in part be driven by temperature differences. Although the PNA concentration is similar in the clean and polluted periods, during the clean period, the average temperature is warmer by 10 K on average. During the polluted period an average temperature of 245 K corresponds to a PNA lifetime against thermal decomposition of approximately 11 h. The PNA lifetime against thermal decomposition reduces to approximately 2 h at 255 K during the warmer, cleaner period. This difference contributes to enhanced thermal decomposition of PNA and a sustained HO₂ concentration at night during the cleaner period.

4. Conclusions

We have explored the HO _x budget of a polluted wintertime Arctic urban environment and how it compares with previous midlatitude wintertime observations and modeling of HO _x . Model-simulated HO _x concentrations in Fairbanks are lower on average than wintertime measurements made in other urban areas such as Beijing. Despite limited solar radiation at short wavelengths and the large solar zenith angle, the photolysis of HONO and formaldehyde occurring at longer wavelengths sustains the oxidizing capacity in Fairbanks. Chemical production alone cannot explain HONO concentration in Fairbanks, and there is uncertainty in HONO emission sources and heterogeneous chemistry, which can make the modeling of OH in the Arctic challenging. − Future work to better understand HONO production in the Arctic could therefore improve the model representation of OH concentrations.

The HO _x budget in our model was impacted by the meteorological conditions in Fairbanks, with strongly stable inversion events having a different atmospheric composition and chemistry from weakly stable periods. For our two case study periods, the strongly stable period was more polluted, and this period had lower HO₂ and O₃ concentrations compared to the weakly stable, clean period. There is large diel variability in the HO₂ concentration throughout our model simulations. The peaks in HO₂ concentration are at opposite times of day in the clean versus polluted periods driven by the different chemical and meteorological regimes. During the clean period, cycling between PNA (HO₂NO₂) and HO₂ + NO₂ is large and contributes to the build-up of HO₂ concentration at night. This may be due in part to the warmer temperatures promoting the thermal decomposition of the PNA. However, this finding is extremely sensitive to assumptions regarding physical controls of the PNA atmospheric residence time. In comparison, nocturnal chemistry is minimal during the polluted period, and there is no accumulation of HO₂ at night. HO₂ production during the day is driven by formaldehyde photolysis and oxidation, especially during the polluted periods. Despite the differences in the HO₂ budget and higher OH reactivity during the polluted period, the OH concentration in the two periods is similar. HONO, and to some extent formaldehyde through HO₂, sustain OH concentration in both periods, and lower NO _x concentrations in the clean period limit OH loss. The largest chemical sink of OH on average is the reaction with NO₂ to form HNO₃, which may be considered a terminal loss of NO _x . However, during the heavily polluted surface-based inversion event from January 29 to February 2, recycling of OH and NO into HONO was the greatest sink of OH. This reaction is not a permanent sink of OH; thus, this reaction may help sustain OH concentrations during pollution events.

When including a measured C₅H₉ fragment as isoprene in the model and adding estimated concentrations of short-chain alkanes and alkenes as constraints, there were significant changes in the HO _x budget. Adding isoprene and alkenes increased loss of OH through reactions with these species but increased HO₂ via multiple production reactions, including decomposition of peroxy radicals produced by OH oxidation. This indicates that DSMACC is sensitive to included VOCs and reveals that the HO _x oxidation of organics is relevant in Arctic urban locations. As there were no measurements of OH concentration or reactivity made during the ALPACA campaign, we cannot directly compare our model results, and our modeling of OH reactivity is limited by measured VOC species. Previous model-observation comparisons of the HO _x budget have reported an underestimate of OH reactivity that may be due to missing VOC measurements. ,,,

Our additional sensitivity analyses demonstrated that model-simulated HO _x is sensitive to the physical loss rate and constrained HONO concentration but not sensitive to relative humidity or uptake. A source of uncertainty comes from our representation of physical loss mechanisms such as mixing. We use a lumped rate to represent the loss through deposition and transport. The midday HO _x concentrations were moderately impacted by modification of the k _phys parameter, but nocturnal chemistry was greatly affected, especially PNA cycling. These results reveal the uncertainty in our model with respect to physical loss and for species that we do not constrain with measured concentrations. Future field studies in Arctic polluted environments that include colocated measurements of PNA, HONO, VOCs, and radical species are needed to corroborate our modeling results.

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsestair.6c00254.

• Comparison between normalized kernel density estimate (KDE) distributions of whole-campaign ALPACA CO measurements and subset of CO measurements used in DSMACC modeling (PDF)

⋈.

Department of Earth and Environmental Sciences, The University of Manchester, Manchester M13 9PL, U.K

⧓.

Center for Atmospheric and Environmental Chemistry, Aerodyne Research Inc., Billerica, Massachusetts 01821, USA

The authors declare no competing financial interest.

Published as part of the ACS ES&T Air special issue “The ALPACA Project - Alaskan Layered Pollution and Chemical Analysis Project”.