Depth-resolved photochemical production of hydrogen peroxide in the global ocean using remotely sensed ocean color

1 Introduction

It is well established that oxidation and reduction (redox) reactions control biogeochemical processes and affect water quality in the environment (Kieber et al., 2003; Morel and Price, 2003; Learman et al., 2011). Early work provided the first theoretical evidence that the redox potential of seawater is controlled exclusively by the oxygen (O₂) system (Cooper, 1937). Since then, a great deal of work has generally confirmed that if present, the O₂ system has essential roles in controlling redox chemistry in the ocean (Keeling et al., 2009; Lu et al., 2018; Zhang et al., 2018). While the oxygen-water redox pair controls redox conditions over long-geological timescales, the situation is more complex, especially on timescales at which many biogeochemical processes occur (Rose, 2016). Along the four-electron pathway between oxygen and water are a set of reactive oxygen species (ROS), namely superoxide (O₂⁻), hydrogen peroxide (H₂O₂) and hydroxyl radical (OH^•), that individually participate in critical redox chemistry (Petasne and Zika, 1987; Sawyer, 1991; Wuttig et al., 2013; Sutherland et al., 2021), and ultimately have significant impacts on the biogeochemical cycling of essential trace elements and reduced carbon in the ocean (Zepp et al., 1992; Scully et al., 2003; White et al., 2003; Rose and Waite, 2006; Rose, 2016). The low seawater concentrations of these ROS (at fM to nM levels) reflect a balance of sources and sinks for each ROS in the system, revealing just how fast the redox cycle spins (Haag and Hoigne, 1986; Mopper and Zhou, 1990; Goldstone and Voelker, 2000).

In marine surface waters, photochemical reactions participate in redox chemistry by creating a diverse blend of oxidized and reduced chemical products (Andrews et al., 2000; Sharpless and Blough, 2014). The absorption of ultraviolet radiation (UVR; 290–400 nm) by chromophoric dissolved organic matter (CDOM) initiates photochemical reactions in the ocean (Mopper et al., 2015). Many of these photochemical reactions involve molecular oxygen, leading to the formation of an array of ROS, including those mentioned above (O₂⁻, H₂O₂, and OH^•) and singlet oxygen (Kieber et al., 2003; White et al., 2003; Rose and Waite, 2006; Kieber et al., 2014; Powers and Miller, 2015a; Ossola et al., 2021). The generation of these redox-active species is ubiquitous in sunlight-irradiated waters containing CDOM, with photochemical ROS production rates dependent on the solar photon flux, absorbance of solar energy, and photochemical reaction efficiency (Cooper and Zika, 1983; Kieber et al., 2014; Powers and Miller, 2014, 2016; Ossola et al., 2021).

In aquatic environments, H₂O₂ is important 1) in the degradation and indirect photobleaching of CDOM and organic pollutants (Goldstone et al., 2002; Scully et al., 2003), 2) in the redox cycling of biologically important trace metals (Moffett and Zika, 1987; Voelker and Sulzberger, 1996; Kieber et al., 2003; Miller et al., 2009), and 3) as an oxidative stress in aquatic ecosystems (Kieber et al., 2003; Zepp et al., 2003; Clark et al., 2008; Morris et al., 2011). H₂O₂ has been widely studied not only due to its biogeochemical reactivity, but also due to its greater concentration and stability relative to other ROS (O’Sullivan et al., 2005). However, it is O₂⁻ that represents the initial electron transfer to O₂ and subsequently serves as the photochemical precursor for H₂O₂. O₂⁻ is especially interesting because it can act as either an oxidant or a reductant, depending on the chemical encounter. If it reacts via an oxidative pathway it returns to the dissolved O₂ pool and again participates in oxidative reactions just like it always did. If it reacts via a reductive pathway, however, it can result in the production of H₂O₂, then OH^• and eventually water with a net “consumption” of O₂ (Sutherland et al., 2020). In this way, the production and reactive pathways for O₂⁻ in seawater serve as crucial controlling factors for many biological and chemical cycles in the photic zone (Rose, 2016).

There is good evidence for widespread dark extracellular O₂⁻ and H₂O₂ production by marine microorganisms (Diaz et al., 2013), confounding the long held idea that photochemistry is the primary marine source of O₂⁻, and therefore, H₂O₂ in sunlit surface waters. A compilation of O₂⁻ production rates by marine cyanobacteria, phytoplankton, and heterotrophic bacteria suggests a marine biological oxygen sink amounting to 15–50% of the net oxygen production (Sutherland et al., 2020). Given that this estimate was not corrected for the much colder waters and slower biological rates in the deep dark ocean, as well as observations that H₂O₂ concentrations are negligible below the surface mixed layer, it is almost certainly an overestimate. However, the contribution of these known biological sources for ROS relative to abiotic photochemical production in sunlit surface waters remains unresolved. Furthermore, measured biological O₂⁻ rates are by definition “net” production rates and are typically determined in the absence of solar radiation (Diaz et al., 2013; Sutherland et al., 2020). It is currently unclear what, if any, changes in observed rates and pathways result from sunrise turning the nighttime into the day, with its myriad of UVR-generated chemical transients. While reduction (via catalase or peroxidase) is typically the dominant pathway for H₂O₂ decay, oxidation pathways have been reported to exceed reduction in some surface coastal waters, suggesting contribution of unknown abiotic or metabolic pathways (Sutherland et al., 2021). Photochemical O₂⁻ measurements may also be flawed, since they are often estimated using steady-state O₂⁻ concentrations in sunlit waters and dark decay rates (Powers et al., 2015). It is clear that decay rates in sunlit waters are not well defined and quantitative data are sorely needed. Thus, our current understanding of these critical reactions required for accurate modeling of the redox state of the surface ocean remains incomplete.

Despite the uncertainties previously discussed, an evaluation of surface ocean photochemical O₂⁻ and H₂O₂ production rates based on the current state of the science is warranted. Powers and Miller (2014) presented a method to estimate global ocean photochemical H₂O₂ and O₂⁻ production using modeled solar irradiances and remotely sensed ocean color and temperature. However, depth-resolved rates were not calculated nor were error estimates presented. In this work, we improve upon this initial method in several ways. Notably the spectral efficiency of H₂O₂ photoproduction used in previous studies, represented by the apparent quantum yield (AQY) spectrum, was previously based on extended irradiations that likely resulted in an underestimate of AQY due to a lack of reciprocity (Powers and Miller, 2014). Here we update H₂O₂ photoproduction models using only AQY spectra determined when H₂O₂ accumulation rates are within a linear range (Kieber et al., 2014). Furthermore, Powers and Miller (2014) used a 2:1 stoichiometric ratio for O₂⁻:H₂O₂ production based on an assumption that O₂⁻ dismutation adequately describes the relationship between O₂⁻ and H₂O₂ in open ocean seawater (Powers and Miller, 2014). Subsequently, direct, simultaneous measurements of O₂⁻ and H₂O₂ photoproduction rates for waters from deep profiles in the Gulf of Alaska revealed an average stoichiometry of 4:1, but with a large range between samples from the surface, 1,000 m, and just above the seafloor (Powers et al., 2015). Most relevant here, the ratio between O₂⁻ and H₂O₂ photoproduction rates for surface waters was highest, averaging 10.1 ± 5.4 or 13.8 ± 6.7 comparing rates normalized to CDOM absorbance (Powers et al., 2015). Work relating the photoproduction of one-electron reductants, the precursors for O₂⁻ and H₂O₂, also suggests that the ratio between O₂⁻ and H₂O₂ photoproduction is well above 2, ranging from 6—16 in a variety of natural and synthetic water samples (Zhang and Blough, 2016; Le Roux et al., 2021). These results are not surprising since the oxidation of O₂⁻ to O₂ is more thermodynamically favorable than reduction to H₂O₂ in conditions present in most surface waters (Sawyer, 1991; Rose, 2016).

Most importantly, photochemical rates in the surface ocean depend on the light available and the absorbance of that light by CDOM. Previous models of oceanic UV attenuation estimated from remotely sensed ocean color have assumed that the fraction of attenuation due to CDOM is 0.68 at 320 nm, a value determined by Fichot and Miller (2010) using a large open ocean data set. However, this average ratio must be extrapolated spectrally and can deviate greatly depending on dissolved water constituents and suspended particles (Cao et al., 2018; Zhu and Kieber, 2020). Algorithms that compute diffuse attenuation coefficients (K_d(λ)) (Fichot et al., 2008; Cao et al., 2014) and CDOM absorption coefficients (a_g(λ)) independently (Cao and Miller, 2015) allow for a more robust estimate of the light available for CDOM-initiated photochemical reactions in the surface ocean. This is especially true in complex inshore waters, where modeled photochemical rates can vary ± 100% between those that use this average a_g(λ)/K_d(λ) ratio and those using a_g(λ) and K_d(λ) determined separately to calculate a specific ratio for each pixel (Cao et al., 2018). We previously assumed that these extreme differences in rates were not a problem for global ocean models (Powers and Miller, 2014), but a_g(λ)/K_d(λ) for UV wavelengths can be much greater than 0.7 (Johannessen, 2003) and recent work demonstrated that modeled a_g(λ)/K_d(λ) ratios exhibited a large range in the open ocean (Zhu and Kieber, 2020). These updates were used along with recent experimental data to provide new constraints on the magnitude of photochemical O₂⁻ and H₂O₂ production in the marine environment.

2 Materials and methods

2.1 Sample collection

Surface seawater (5 m) samples were collected at three stations in the South Atlantic Bight (SAB) aboard the R/V Savannah in March 2018. The geographic locations of the sampling stations are depicted in Figure 1. All samples were 0.2 μm gravity-filtered (Whatman Polycap 36AS) directly from Niskin bottles into combusted (4 h at 500°C) 1 L borosilicate glass bottles (Kimax). All tubing, filters, and bottle caps were pre-cleaned with dilute (0.1 M) HCl and rinsed several times with 18.2 MΩ cm Milli-Q water (Millipore). Bottles were rinsed three times with filtered sample before final collection. Samples were irradiated within hours of collection, or within 1–2 days prior to an experiment. Samples not irradiated immediately after collection were stored at 4°C until use. Because many H₂O₂ photoproduction rates and AQY measurements have been made on stored samples (Kieber et al., 2014; Powers and Miller, 2014), shipboard irradiations were also conducted to determine H₂O₂ time-dependent photoproduction rates using freshly collected samples. These experiments were also used to assess how well modeled photoproduction rates predicted measured rates in freshly collected samples.

FIGURE 1

FIGURE 1. Locations of the sampling stations in the South Atlantic Bight. Bathymetry and topography data from the Global Self-consistent, Hierarchical, High-resolution Geography Database (GSHHG) were plotted using the m_map package for Matlab® (Pawlowicz, 2020). Water-column color coding indicates the depth in meters.

2.2 Laboratory H₂O₂ determination

H₂O₂ concentrations in seawater samples (including irradiated samples and dark controls) were measured with a flow injection analysis (FIA) chemiluminescence (CL) system (Waterville Analytical, FeLume) using methods described in detail previously (King et al., 2007; Powers et al., 2015). Briefly, the carrier solution (0.2 M HCl) and chemiluminescent reagent, 2 μM 10-methyl-9-(p-formylphenyl)acridinium carboxylate trifluoromethanesulfonate (AE), were mixed and pumped to a spiral flow cell (Global FIA) using a peristaltic pump (Rainin). A ∼0.2 ml sample loop was filled by syringe and the sample was introduced to the carrier stream using a 10-port valve (VICI). The sample and carrier–AE solution mixed with 0.1 M carbonate buffer (pH 10.4) to initiate CL in the flow cell, which was situated below a photomultiplier tube (PMT; Hamamatsu HC135 PMT, 900 V, 400 ms integration). The PMT signal was recorded using Waterville Analytical software (LabView); 3—5 injections were repeated per sample. The FIA system was calibrated with H₂O₂ standards prepared in seawater using a 2 mM H₂O₂ stock solution diluted from 30% H₂O₂ (J.T. Baker), with concentrations confirmed using its absorbance at 240 nm and a molar absorptivity of 38.1 M⁻¹ cm⁻¹ (Miller and Kester, 1988).

2.3 Photochemical experiments

Photochemical experiments were conducted onboard the R/V Savannah with a Suntest CPS solar simulator (Atlas) equipped with a 1.5 kW Xenon lamp and a daylight filter to limit irradiation below 290 nm. These photochemical experiments were conducted to test for linearity in H₂O₂ photoproduction and to compare rates measured using freshly collected samples with those modeled as described in Sections 3.2, 3.4. Prior to irradiation, seawater samples were partitioned into 15 matched 10 cm cylindrical quartz spectrophotometric cells (Spectrocell) at room temperature. Cells were filled with no headspace and sealed using gas-tight screw caps with Teflon-faced butyl rubber septa. To ensure reproducible and quantifiable light exposures, cells were oriented vertically directly below the solar simulator in a black, anodized, water-cooled aluminum irradiation block. This block allowed no transfer of light between cells. The block was fitted with a half-inch thick grey PVC lid with 1.6 cm diameter holes centered directly above each cell window to remove residual off-axis photons and maximize collimated radiation within the cell, thus ensuring a reproducible and known pathlength for calculations (Powers and Miller, 2015b). The block was maintained at 20°C using a circulating water bath and cells were removed at various time points throughout each experiment to measure H₂O₂ and determine rates. One dark control was kept in the block and shielded from light while all other dark controls were kept in the water bath.

The spectral downwelling irradiance ($E_{0,lab}\left(\lambda \right)$; mol photons m⁻² s⁻¹ nm⁻¹) entering each cell was measured below the PVC lid for each cell position at 1 nm intervals with a UV-Vis portable spectroradiometer (OL756; Optronic Laboratories). The OL756, fitted with a quartz fiber optic cable and 2-inch diameter integrating sphere, was calibrated with a NIST standard lamp (OL752-10 irradiance standard; Optronic Laboratories) pre and post cruise. Under the irradiation conditions described above, the average integrated UV photon dose (290–400 nm) entering each cell was 33 W m⁻² (∼8.72 mol photons m⁻² d⁻¹). The sample absorbance ($A\left(\lambda \right)$), before and after irradiation, was measured directly in each 10-cm cell using a single-beam UV-Vis spectrophotometer (Agilent). Milli-Q water was used as the absorbance blank and each $A\left(\lambda \right)$ sample spectrum was corrected for offsets or instrument drift by subtracting the average absorbance value in the 680–720 nm range. Corrected $A\left(\lambda \right)$ were converted to Naperian absorption coefficients ($a_{g,lab}\left(\lambda \right)$; m⁻¹) with the following equation:

$a_{g,lab}\left(\lambda \right)=2.303\times A\left(\lambda \right)/L(1)$

where $L$ is the cell pathlength (0.1 m). For these laboratory irradiations, H₂O₂ photoproduction rates ($P_{H_2{O}_2}\left(\lambda \right)$) were estimated using measured $a_{g,lab}\left(\lambda \right)$, $E_{0,lab}\left(\lambda \right)$, and the marine-average ${\Phi }_{H_2{O}_2,293K}\left(\lambda \right)$ with the following equation:

$P_{H_2{O}_2}\left(\lambda \right)=E_{0,lab}\left(\lambda \right)\times S\times {\Phi }_{H_2{O}_2,293K}\left(\lambda \right)\times \left(1-e^{-a_{g,lab}\left(\lambda \right)\times L}\right)/V(2)$

where S is the irradiated surface area (m²) and $V$ is the volume (m³). ${\Phi }_{H_2{O}_2,293K}\left(\lambda \right)$ was calculated using Eqs. 5, 6 below. Simulated production rates were then determined by integrating Eq. 2 from 290—400 nm. The integrated photon dose ($Q_{a\left(int\right)}$) absorbed by CDOM from 290—400 nm for a given irradiation time interval was also calculated using Eq. 3.

$Q_{a\left(int\right)}={\int }_{290}^{400}{E}_{0,lab}\left(\lambda \right)\times S\times \left(1-e^{-a_{g,lab}\left(\lambda \right)\times L}\right)d\lambda (3)$

2.4 Global photochemical model

In general, we followed the methods outlined in Fichot and Miller (2010) and Zhu and Kieber (2020) for global photochemical flux estimates. Briefly, H₂O₂ photochemical production rates ($P_{H_2{O}_2,z}$; mol m⁻³ d⁻¹) were estimated at each depth (z; m) with Eq. 4:

$P_{H_2{O}_2,z}={\int }_{290}^{400}{{{\Phi }_{H_2{O}_2,T}\left(\lambda \right)\times a}_g\left(\lambda \right)\times E}_{{CL,0}^{-}}\left(\lambda \right)\times e^{-K_d\left(\lambda \right)\times z}d\lambda (4)$

where ${\Phi }_T\left(\lambda \right)$ is the temperature-corrected AQY {[mol(mol photons)⁻¹]}, $a_g\left(\lambda \right)$ is the satellite-retrieved absorption coefficient of seawater (m⁻¹), $E_{{CL,0}^{-}}\left(\lambda \right)$ is the modeled, cloud-corrected, downwelling scalar irradiance immediately below the air-seawater interface (mol photons m⁻² d⁻¹ nm⁻¹), and $K_d\left(\lambda \right)$ is the satellite-retrieved diffuse attenuation coefficient for downwelling irradiance (m⁻¹). All global photochemical flux estimates reported in this work were conducted with a spectral resolution of 1 nm and a vertical depth (z) resolution of 0.1 m. We followed the procedure outlined in Zhu and Kieber (2020) to calculate $P_{H_2{O}_2,z}$ in the surface mixed layer, since the photochemical production integrated over this layer is expected to comprise 94–96% of all the photoproduction in the sunlit layer (Fichot and Miller, 2010). The mixed layer depth (MLD) data were obtained from the Monthly Isopycnal/Mixed-layer Ocean Climatology (https://www.pmel.noaa.gov/mimoc/, Schmidtko et al. (2013)).

We followed Zhu and Kieber (2020) and Cao et al. (2018) for the remote sensing retrievals of $a_g\left(\lambda \right)$ and $K_d\left(\lambda \right)$. Briefly, monthly climatologies of MODIS-Aqua reflectance ($R_{rs}\left(\lambda \right)$) from 2007 to 2017 were downloaded from the NASA Goddard Earth Sciences Data and Information Services Center (GES DISC, https://disc.gsfc.nasa.gov/). The monthly averaged $R_{rs}\left(\lambda \right)$ data, originally observed at MODIS wavebands ($\lambda$ =412, 443, 469, 488, 531, 547, 555, 645, 667 and 678 nm), were linearly interpolated to SeaWiFs wavebands ($\lambda$ =412, 443, 490, 510, 555, and 670 nm) prior to algorithm application. This was done to match SeaCDOM (Cao and Miller, 2015) and SeaUV (Fichot et al., 2008; Cao et al., 2014) algorithms that were subsequently used to retrieve $a_g\left(\lambda \right)$ and $K_d\left(\lambda \right)$, respectively. Note that this work used SeaUV with the updated parameters for offshore waters provided in Cao et al. (2014) to retrieve $K_d\left(\lambda \right)$ at 320, 340, 380, 412, 443, and 490 nm. To extend $K_d\left(\lambda \right)$ retrieved at six wavelengths to the full UV spectrum, $K_d$ (320–400 nm) was determined with a cubic interpolation and $K_d$ (290–319 nm) was calculated with a spectral slope coefficient determined between $K_d$ (320 nm) and $K_d$ (340 nm). The final data product of remote sensing retrievals of $a_g\left(\lambda \right)$ and $K_d\left(\lambda \right)$ were obtained with a wavelength resolution of 1 nm and a spatial resolution of 1° × 1°.

The SMARTS 2.9.5 model (Gueymard, 2001) was used to calculate monthly climatologies of the daily integrated spectral solar irradiance just above the sea surface under clear skies ($E_{0^{+}}\left(\lambda \right)$) from 2007 to 2017. Details regarding the computation of $E_{0^{+}}\left(\lambda \right)$, and the conversions from $E_{0^{+}}\left(\lambda \right)$ to clear-sky downwelling scalar irradiance immediately below the air-seawater interface $E_{{clear,0}^{-}}\left(\lambda \right)$ and from $E_{{clear,0}^{-}}\left(\lambda \right)$ to $E_{{CL,0}^{-}}\left(\lambda \right)$ are provided in Zhu and Kieber (2020). The loss of radiation at the air-seawater interface was calculated following the procedure outlined in Galí et al. (2016). A cloud correction factor (CF) was computed using the cloud-corrected and clear-sky UV spectral irradiance data at 308, 310, 324, and 380 nm, which were obtained from OMI/Aura Surface UVB Irradiance and Erythemal Dose dataset (downloaded from GES DISC). Details of the OMI UV products are provided by Tanskanen et al. (2006). Modeling $E_{0^{+}}\left(\lambda \right)$, $E_{{clear,0}^{-}}\left(\lambda \right)$, and $E_{{CL,0}^{-}}\left(\lambda \right)$ were performed with a wavelength resolution of 1 nm and a spatial resolution of 5 ° × 5 °. Cubic interpolation was applied to $E_{{CL,0}^{-}}\left(\lambda \right)$ to obtain a final data product with a spatial resolution of 1° × 1°.

For H₂O₂ photoproduction, Kieber et al. (2014) compared AQYs from a variety of seawater sources, including coastal and oligotrophic stations in Antarctica, the Pacific Ocean at Station ALOHA, the Gulf of Mexico, and several sites along the eastern coast of the United States. They found that AQYs exhibited limited variability for all regions studied when corrected for temperature. For our H₂O₂ photochemical production estimates, the following equation was used to estimate wavelength-dependent AQYs at 25°C:

${\Phi }_{H_2{O}_2,298K}\left(\lambda \right)=1.70\times e^{-0.0272\times \lambda }(5)$

The AQY spectrum in Eq. 5 was obtained from non-linear regression analyses of combined AQY data from all seawater sources studied by Kieber et al. (2014).

Temperature-dependent AQY (${\Phi }_{H_2{O}_2,T}\left(\lambda \right)$) were calculated using an activation energy of 21.8 kJ mol⁻¹, as suggested by Kieber et al. (2014):

$\mathit{ln}\frac{{\Phi }_{H_2{O}_2,T}\left(\lambda \right)}{{\Phi }_{H_2{O}_2,298K}\left(\lambda \right)}=\frac{21.8}{R}\left(\frac{1}{298}-\frac{1}{T}\right)(6)$

where R is the universal gas constant (8.314 × 10^–3 kJ mol⁻¹ K⁻¹) and T is the temperature (K). For our global estimate, ${\Phi }_{H_2{O}_2,T}\left(\lambda \right)$ appropriate for each pixel was adjusted using 11-year (2007–2017) averaged sea surface temperature (SST) using MODIS Aqua monthly datasets downloaded from GES DISC.

3 Results

3.1 H₂O₂ photoproduction in the South Atlantic Bight

H₂O₂ photoproduction rates were determined from the differences between H₂O₂ concentrations in irradiated seawater samples and dark controls. Changes in photoproduced [H₂O₂] were examined during prolonged irradiations of filtered seawater collected from three stations in the SAB (Figure 2). These samples, designated as Altamaha Outflow, Coastal, and Offshore Stations provided a gradient from colored inshore to blue offshore water samples for the shipboard irradiations used to test our model’s accuracy in different CDOM domains. These samples had average $a_{g,lab}$ (325 nm) values of 2.13, 0.56, and 0.22 m⁻¹, respectively. $Q_{a\left(int\right)}$ (Eq. 3) for the final time points were 90 µmol photons (12 h) for the Altamaha Outflow Station, 75 µmol photons (20 h) for the Coastal Station and 37 µmol photons (48 h) for the Offshore Station. However, H₂O₂ rates were no longer linear at these final timepoints, and therefore integrated photon doses calculated in this way do not have a simple, constant relationship to H₂O₂ photoproduction.

FIGURE 2

FIGURE 2. Photoproduced H₂O₂ versus irradiation time (h) for samples collected in the South Atlantic Bight in March 2018.

Previously, it was shown that the photoproduction of H₂O₂, as well as several other CDOM photoproducts (including CO, CO₂, and carbonyl compounds), is a nonlinear function of the light exposure time (Kieber et al., 2014; Powers and Miller, 2015b; Zhu and Kieber, 2018). Similarly, nonlinear accumulation was observed here for H₂O₂ photoproduction in seawater from the Altamaha Outflow, Coastal, and Offshore Stations (Figure 2), occurring at $Q_{a\left(int\right)}$ values of ∼ 45, 12.5, and 9 µmol photons, respectively. Furthermore, H₂O₂ production rates calculated from a linear regression of [H₂O₂] with time, using all time points, were ∼ 46%, 40%, and 65% lower, respectively, than H₂O₂ production rates determined before nonlinearity occurred. This result agrees with Kieber et al. (2014) who also reported a lack of reciprocity in H₂O₂ photoproduction during monochromatic irradiations (300–400 nm) of Gulf of Maine seawater. It should be noted here that nonlinearity cannot be accounted for by CDOM fading and that the photon dose at which nonlinear accumulation begins is sample dependent and should be tested prior to any AQY determination.

Under natural sunlight, the transition to non-linear H₂O₂ photoproduction should occur only after 1–2 days of exposure or more. This is because the daily downwelling solar irradiation in mid-latitude regions of NW Atlantic entering the ocean is ∼ three times less than the daily-integrated irradiation output of the 1.5 kW Xenon lamp used in our shipboard irradiations. Furthermore, downwelling UV irradiance decreases exponentially with depth (Fichot and Miller, 2010; Zhu and Kieber, 2019). Depending on CDOM, MLD, and surface mixing rates, the average $Q_{a\left(int\right)}$ for any chromophore mixed through this varied radiation field is far smaller than the factor of three difference in initial $E_{{CL,0}^{-}}\left(\lambda \right)$ from simulated irradiation. This fact also argues that when extrapolating lab-determined photoproduction rates to surface ocean environments, AQYs determined from production rates over the initial linear range are most appropriate for accurate prediction. Consequently, the wavelength- and temperature-dependent AQYs used in our photochemical estimates were obtained using relatively short irradiation times (ca. 10–240 min depending on the seawater absorbance, the selected wavelength, and temperature; Kieber et al., 2014). The global photochemical estimates presented here (Sections 3.4, 3.5), using the updated AQY spectrum calculated from Eq. 5, are thus an improvement of the global AQY spectrum presented in Powers and Miller (2014), which used proper laboratory exposures but it was just that the time was wrong. In that work, irradiations lasted ∼12 h, corresponding to 27 µmol photons absorbed under UV light treatments, which was well beyond the 9 µmol photon dose where nonlinearity was observed for the SAB Offshore Station (Figure 2).

3.2 Natural variability and errors associated with AQYs

For H₂O₂ photoproduction, a single “marine-average” AQY spectrum (Eq. 5) derived by Kieber et al. (2014) from a large variety of seawater sources, was used to model global photochemical production rates. As a limited test for its general applicability, we used this same average AQY to predict H₂O₂ production rates in our SAB samples. All wavelength-dependent AQY data reported by Kieber et al. (2014) and the “marine-average” AQY spectrum are shown in Figure 3A. The “marine-average” AQY spectrum is not statistically different from the AQY spectrum obtained using pooled AQY data in the Caribbean Sea, Orinoco River outflow (Venezuela), Suwanee River (Georgia, United States), Vineyard Sound (Massachusetts, United States), Shark River outflow (Florida, United States), and the Antarctic (Yocis et al., 2000). Here we also compare the “marine-average” AQY to the AQY spectra measured for the SAB by Powers and Miller (2016) (Figure 3A). For direct comparison with Powers and Miller (2016), the Kieber et al. (2014) “marine-average” AQY spectrum was fit to the following equation:

${\Phi }_{H_2{O}_2,298K}\left(\lambda \right)=e^{-\left(m_1+m_2\left(\lambda -290\right)\right)}(7)$

where m₁ and m₂ are the fitting parameters obtained by fitting AQY data to Eq. 7. The fitting parameters with the 95% CI for AQY reported in Kieber et al. (2014) and Powers and Miller (2016) are presented in Table 1. As seen in Table 1, m₂ from the two studies are not significantly different, given the overlap in the 95% CI for m₂. A significant difference was observed for m₁ after a temperature correction (from 20°C to 25°C) was applied to the Powers and Miller (2016) AQY data. The difference in m₁ translates to a 23% difference in ${\Phi }_{H_2{O}_2,298K}$ at 290 nm. An underestimation in AQYs reported by Powers and Miller (2016) is expected given that the AQY spectra were determined from longer irradiations, which were likely beyond the initial linear range of [H₂O₂] accumulation over time. Given that the difference in m₁ is smaller than the measurement uncertainties associated with the determination of wavelength-dependent AQYs (Table 2), the observed similarity in AQY spectra from different seawater sources with various optical properties gives us confidence that a single AQY equation (i.e., Eq. 5) can be used in our global photochemical model.

FIGURE 3

FIGURE 3. (A) “Marine-average” AQY spectrum [Φ(λ)] versus wavelength (λ; nm) (solid black line) computed from all measured monochromatic AQY data at 25°C reported in Kieber et al. (2014) (black open dots). Dashed black lines are the modeled AQY spectrum calculated using upper and lower limits of m₁ and m₂ in Table 1. The solid blue line shows pooled Φ(λ) computed from polychromatic AQY data reported in Powers and Miller (2016). Φ(λ) from individual experiments reported in Powers and Miller (2016) are depicted with blue dotted lines. (B) Simulated H₂O₂ photoproduction rates ($P_{H_2{O}_2}\left(\lambda \right)$) calculated for the Offshore Station using Eq. 2. The solid black line represents $P_{H_2{O}_2}\left(\lambda \right)$ estimated with the “marine-average” AQY spectrum, and the blue squares represent $P_{H_2{O}_2}\left(\lambda \right)$ calculated with average AQY values at each wavelength (error bars represent 95% CI). Dashed black lines represent $P_{H_2{O}_2}\left(\lambda \right)$ with ±13% error added.

TABLE 1

TABLE 1. Comparison between AQY parameters in Powers and Miller (2016) and the “marine-average” AQY parameters used in the global model. For monochromatic irradiations, AQY parameters were derived by a nonlinear regression of AQY versus λ. Uncertainties listed for m₁ and m₂ represent the margin of error with 95% confidence level. For polychromatic data, the parameters were determined using a non-linear curve-fitting routine in Matlab^® (nlinfit), to solve for the best-fit equation to describe all light treatment data.

TABLE 2

TABLE 2. Summary of marine H₂O₂ Φ(λ) data [×10^–4 mol/(mol photons)] determined at 25°C (Kieber et al., 2014) ± 1 SD, when available. ND is not determined.

To assess the natural variability associated with wavelength-dependent AQYs, a summary of measured ${\Phi }_{H_2{O}_2,298K}\left(\lambda \right)$ at 290, 300, 320, 340, 360, and 400 nm (originally reported in Kieber et al., 2014) is presented in Table 2. The standard deviation in ${\Phi }_{H_2{O}_2,298K}\left(\lambda \right)$ increased with increasing wavelength, from 27% at 290 nm to 37% at 400 nm. There are more data points available in the UVB (290–320 nm) than at longer wavelengths, perhaps driving some of these increases. The Offshore Station in the SAB from the current study was used to examine how $P_{H_2{O}_2}\left(\lambda \right)$ (calculated from Eq. 2) varied as a function of wavelength (Figure 3B). The average AQYs at 290, 300, 320, 360, and 400 nm (Table 2) were used along with the average $E_{0,lab}\left(\lambda \right)$ and $a_{g,lab}\left(\lambda \right)$ at these wavelengths to calculate wavelength-dependent H₂O₂ photochemical production rates in the Offshore Station sample (data shown in blue squares, Figure 3B). The $P_{H_2{O}_2}\left(\lambda \right)$ calculated using the “marine-average” AQY is in reasonably good agreement with that calculated using wavelength-dependent AQY values reported in Kieber et al. (2014) with H₂O₂ photoproduction peaking at ∼320 nm. Error (95% confidence level) in the “marine-average” AQY slope (i.e., m₂) is ∼ ±11% (Table 1), resulting in a ±13% change in the integrated photoproduction rate from 290 to 400 nm. As seen in Table 2, the SD of measured AQYs at 320 nm is 36%, higher than the percentage differences between the predicted and measured rates as discussed below (Table 3). This measurement error likely drives most of the error in modeled photoproduction rates.

TABLE 3

TABLE 3. Predicted H₂O₂ photoproduction rates in the quartz irradiation cells using the “marine-average” AQY and measured H₂O₂ photoproduction rates (± 1 SD) in filtered water samples at two time points. % differences between measured and predicted rates are also listed.

To further justify the use of the “marine-average” AQY in our global estimates, we compared measured H₂O₂ photoproduction rates in our SAB samples to those predicted using the “marine-average” AQY (${\Phi }_{H_2{O}_2,293K}\left(\lambda \right)$) with measured $E_{0,lab}\left(\lambda \right)$ and $a_{g,lab}\left(\lambda \right)$ values from the short-term irradiations according to Eq. 2. Predicted production rates (calculated by integrating Eq. 2 from 290—400 nm) are presented in Table 3 along with measured rates and the % differences between the measured and predicted rates. All calculated production rates are in good agreement with measured rates (±9.4–27% in coastal to offshore water; Table 3). The biggest differences were observed for the Altamaha Outflow sample, and for samples that were exposed to prolonged irradiation with non-linear relationships between [H₂O₂] and photon exposure (e.g., the offshore sample exhibited a 15% error in predicted photoproduced H₂O₂ after 5 h versus 27% error in predicted photoproduced H₂O₂ after 12 h). Based on data shown in Table 3, we suggest that the “marine-average” AQY adequately represents the overall efficiency of H₂O₂ photoproduction in the SAB, with percentage differences between predicted and measured rates < 33%.

3.3 Errors associated with apparent quantum yield temperature adjustments using E_a

Natural variability was observed when fitting the Arrhenius equation to H₂O₂ data used to determine $E_a$ (kJ mol⁻¹):

$\mathit{ln}\mathrm{\Phi }\left(\lambda \right)=\mathit{ln}{A}_p-\frac{E_a}{RT}(8)$

where A_p is a pre-exponential factor, R is the universal gas constant (8.314 × 10^–3 kJ mol⁻¹ K⁻¹), and T is temperature (K). The $E_a$ and r² derived by fitting temperature-dependent AQYs in Kieber et al. (2014) to Eq. 8 are presented in Table 4. The resulting average $E_a$ and 95% CI determined from the slope of Arrhenius equation (Eq. 8) for each wavelength is shown in Figure 4A, and the corresponding Arrhenius plot for all AQY data from 0—35°C (gray dots) reported in Kieber et al. (2014) are presented in Figure 4B. Based on a pooled linear regression (Figure 4B), overall error in E_a(λ) was ±32%, but it was only ±20% when only wavelengths less than 360 nm were considered (Table 4). It should be noted that in laboratory irradiations, 82% (coastal water) to 86% (open ocean water) of the H₂O₂ photoproduction under UVR occurred at λ < 360 nm. Given this wavelength dependence, $E_a$ used here is likely to generate errors closer to 20%. The reader should note that errors associated with $E_a$ do not translate linearly to errors in AQY. For example, there was only a 5% difference in ${\Phi }_{H_2{O}_2,303K}$ when a 30% error in $E_a$ was introduced into its 5°C adjustment from ${\Phi }_{H_2{O}_2,298K}$ using Eq. 6. Consequently, errors contained in using a single “marine average” AQY for model calculations are generally larger than any correction for temperature.

TABLE 4

TABLE 4. $E_a$ (kJ mol⁻¹) determined from the slope of the linear regression of ln [Φ(λ)] versus 1/T (K) and an R value of 8.315 × 10^–3 kJ K⁻¹ mol⁻¹ at each wavelength as listed. Data were originally reported by Kieber et al. (2014). A pooled $E_a$ was determined by a regression of all ln [Φ(λ)] versus 1/T data, regardless of wavelength. $E_a$ at λ < 360 and λ ≥ 360 were also determined to assess the variability in $E_a$ used in the global H₂O₂ model. Errors associated with $E_a$ denote the margin of error with 95% confidence level.

FIGURE 4

FIGURE 4. (A) Average activation energies ($E_a$) and 95% CI determined from the slope of the linear regression ln [Φ(λ)] versus 1/T (K⁻¹) for AQY data at each wavelength. (B) ln (AQY) versus 1,000/T (K⁻¹) (best-fit line is shown in solid black line; 95% prediction interval is bracketed by dashed lines) for all AQY data at from 0—35°C (gray dots) reported in Kieber et al. (2014).

3.4 Modeled depth-dependent H₂O₂ photoproduction

Monthly climatologies of H₂O₂ photochemical production rates were determined at z = 0, 1, and 20 m using Eq. 4. Corresponding global maps of computed photoproduction rates are presented in Figure 5. We show global maps for February, May, August, and November, as representatives for the spatial variability of H₂O₂ photoproduction rates during each season. Global maps for $E_{{CL,0}^{-}}$ (320 nm), $a_g$ (320 nm), $K_d$ (320 nm), and SST in February, May, August, and November are presented in the Supplementary Material, Supplementary Figures S1–S4.

FIGURE 5

FIGURE 5. Calculated H₂O₂ photoproduction rates at (A) 0 m, (B) 1 m, and (C) 20 m in February, May, August and November. Note the difference in scale between (A–C). White indicates no data.

In surface waters, the spatial and vertical distribution of the modeled H₂O₂ photoproduction rate is similar to that observed for carbonyl compounds (Zhu and Kieber, 2020). In particular, the spatial distribution was mainly controlled by UV irradiance, the CDOM absorption coefficient and seawater temperature. High CDOM-absorbing coastal and upwelling areas exhibited the highest H₂O₂ photoproduction rates at 0 and 1 m (Figure 5). Equatorial waters also exhibited relatively high H₂O₂ photoproduction rates among open-ocean waters, due to their relatively high SST, upwelling in the equatorial eastern Pacific and the equatorial eastern Atlantic, as well as the high abundance of UV exposure. At mid-to high-latitudes, higher H₂O₂ photoproduction rates were found during the summer (e.g., August in the Northern Hemisphere) as compared to the winter (e.g., February in the Northern Hemisphere). The vertical distribution of H₂O₂ photoproduction rate in the water column was controlled by the exponential attenuation of UV radiation in the water column, resulting in H₂O₂ photoproduction rates at 20 m that were ∼1–2 orders of magnitude lower than rates computed for the same location at 0 m. As expected, higher H₂O₂ photoproduction rates were calculated for oligotrophic waters at 20 m compared to coastal and upwelling regions, due to deeper penetration of UV radiation compared to high CDOM-absorbing waters (Supplementary Figure S3).

Monthly 11-year (2007–2017) averaged estimates of H₂O₂ photoproduction rates at 0 m (computed with Eq. 4) are compared to previous literature reports of in situ H₂O₂ photoproduction rates to further evaluate our global model. Here we focus on low CDOM-absorbing waters in the open ocean that cover the majority of the surface area of the global oceans. Kieber et al. (2014) reported a daily H₂O₂ photoproduction rate of 72 nM d⁻¹ at an oligotrophic station near 40 °N 70 °W in July 1999. By comparison, our modeled H₂O₂ photoproduction rate at 0 m for this location was 64 nM d⁻¹ in July, only a 12% difference between the measured and climatologically modeled rate using an average marine AQY. This is within the ±33% error previously determined comparing our measured and predicted H₂O₂ photoproduction rates (Section 3.2). In Antarctic waters, Yocis et al. (2000) reported 2.4–3.5 nM h⁻¹ in situ photoproduction rates near 60° S 55° W in summer, which translates to ∼24–35 nM d⁻¹ assuming that the daily integrated photon flux equals ten times the hourly photon flux measured during the incubation experiment. Our modeled rate for this location is 17 nM d⁻¹ in December and is ∼29% lower than the measured range. At our Offshore Station (near 30° N 79° W) in the SAB, we measured a 3.7 nM h⁻¹ H₂O₂ photoproduction rate in March using a solar simulator with an integrated photon flux of 0.36 mol photons m⁻² h⁻¹. For this location, integrating $E_{{CL,0}^{-}}\left(\lambda \right)$ from 290 to 400 nm yielded an integrated photon flux of 3.2 mol photons m⁻² d⁻¹ in March. Adjusting our measured H₂O₂ hourly rate to the modeled daily photon flux results in a value of 32.9 nM d⁻¹, higher than the 22.4 nM d⁻¹ rate we modeled for 0 m. Even when forcing this comparison between laboratory solar simulation (with spectra from an artificial light source) and a 11-year climatological estimate based on natural sunlight, the ∼32% difference between the measured and modeled photoproduction is not so far away and within the ±36% error that is associated with the natural variability of AQYs at 320 nm (Table 2).

A reasonable agreement between modeled and measured H₂O₂ photoproduction rates was also reported by Powers and Miller, (2014). The estimated H₂O₂ photoproduction rate at 1 m from this previous modeling effort was 63 nM d⁻¹ at 40 °N 70 °W in July, and 32.5 nM d⁻¹ at 55° S 55° W in December. In the present study, our modeled H₂O₂ photoproduction rates at 1 m isodepth are 49.4 nM d⁻¹ at 40 °N 70 °W in July, and are 24.4 nM d⁻¹ at 55° S 55° W in December. The ∼30% difference between estimated photoproduction rates from the two models is within the ±36% error that is associated with the natural variability of AQYs at 320 nm (Table 2).

3.5 Depth-integrated H₂O₂ photoproduction fluxes

Global maps depicting depth-integrated H₂O₂ photoproduction fluxes in the surface mixed layer ($P_{H_2{O}_2,integ}$) were calculated by integrating $P_{H_2{O}_2,z}$ (Eq. 4) from the ocean surface to the bottom of the surface mixed layer. Monthly climatologies for February, May, August, and November were used as representatives for each season in Figure 6. In the water column, ∼96% of UV radiation is absorbed by CDOM in the surface mixed layer (Fichot and Miller, 2010) with an exponential decrease in UV intensity with depth in this layer; this is the most important factor affecting the geographical distribution of $P_{H_2{O}_2,integ}$. Thus, CDOM is the main driver for H₂O₂ photoproduction in our model, which is not surprising since it is inherent in the very concept of AQY used in this, and essentially all other environmental photochemical studies of natural processes. Given that the AQY is experimentally defined as the molar ratio of product (H₂O₂) and photons absorbed (CDOM), the partitioning of UV radiation between CDOM absorption and attenuation by other factors is the critical element in defining photochemical processes with AQY spectra. While there is a potential contribution of H₂O₂ photoproduction from the particulate phase, we have no data to support this so it is neglected in our model. The good agreement between CDOM modeled and measured in situ photochemical rates argues that, at least for H₂O₂, omitting particulate photoproduction is acceptable. Therefore, the $a_g$ to $K_d$ ratio (i.e., the ratio between CDOM absorption and total light attenuation) is the critical optical relationship needed to predict global patterns in $P_{H_2{O}_2,integ}$. As seen in Figure 6A, the highest $P_{H_2{O}_2,integ}$ are found in equatorial waters, as these waters receive a high abundance of UV radiation (Supplementary Figure S1) and exhibit relatively high $a_g$ to $K_d$ ratios (Figure 6B) and SST (Supplementary Figure S4).

FIGURE 6

FIGURE 6. (A) Global maps of $P_{H_2{O}_2,integ}$, (B) $a_g$ to $K_d$ ratios at 320 nm, (C) $P_{H_2{O}_2,integ}^{\prime }$, (D) the percentage difference between $P_{H_2{O}_2,integ}^{\prime }$ and $P_{H_2{O}_2,integ}$ calculated from $\frac{{P_{H_2{O}_2,integ}^{\prime }-P}_{H_2{O}_2,integ}}{P_{H_2{O}_2,integ}}\times 100$. Data in February, May, August and November are displayed. White color in panel (A–C) and black color in panel (D) indicates no data.

To directly compare our photochemical model with several previous studies that used a constant $a_g$ to $K_d$ ratio in their global photoproduction models, we applied this same approach to estimate depth-integrated H₂O₂ photoproduction ($P_{H_2{O}_2,integ}^{\prime }$) in sunlit waters without the capacity to adjust a_g/K_d based on ocean color data as follows:

$P_{H_2{O}_2,integ}^{\prime }={\iint }_{290}^{400}{{\Phi }_{H_2{O}_2,T}\left(\lambda \right)\times a_g^{\prime }\left(\lambda \right)\times E}_{{CL,0}^{-}}\left(\lambda \right)\times e^{-K_d^{\prime }\left(\lambda \right)\times z}d\lambda dz(9)$

where $a_g^{\prime }\left(\lambda \right)$ was computed following Fichot and Miller (2010) using the original SeaUV algorithm (Fichot et al., 2008) to estimate $K_d^{\prime }\left(\lambda \right)$, a constant $a_g$ to $K_d$ ratio of 0.68 at 320 nm, and an assumed exponential function: $a_g^{\prime }\left(\lambda \right)=a_g^{\prime }\left(320\right)\times e^{-S_{a_g}\times \left(\lambda -320\right)}$ ($S_{a_g}$ = 0.0194) to derive a CDOM spectrum for photochemical production. Although depth-integrated photochemical production rates were not calculated in Powers in Miller (2014), the same $a_g^{\prime }\left(\lambda \right)$ calculation was applied in that earlier work. Our calculated $P_{H_2{O}_2,integ}^{\prime }$ in February, May, August, and November are displayed in the global maps in Figure 6C. The percentage difference between $P_{H_2{O}_2,integ}$ and $P_{H_2{O}_2,integ}^{\prime }$ are mapped in Figure 6D. $P_{H_2{O}_2,integ}^{\prime }$ were significantly higher than $P_{H_2{O}_2,integ}$ in coastal and upwelling regions, which represent relatively low a_g/K_d (marked with deep blue in Figure 6B). We also found that regions with a_g/K_d (320 nm) > 0.8 (marked with yellow in Figure 6B) exhibited lower values of $P_{H_2{O}_2,integ}^{\prime }$ compared to $P_{H_2{O}_2,integ}$ (Figure 6D). In open ocean areas where a_g/K_d (320 nm) was close to 0.68 (i.e., the assumed ratio used to calculate $a_g^{\prime }\left(320\right)$), $P_{H_2{O}_2,integ}^{\prime }$ was only slightly higher than $P_{H_2{O}_2,integ}$. The assumed $S_{a_g}$ of 0.0194 may have contributed to overestimation of $P_{H_2{O}_2,integ}^{\prime }$ in open-ocean areas, since this ratio was determined by pooling $S_{a_g}$ data from a large variety of seawater sources including CDOM-rich, productive seawaters and oligotrophic waters. $S_{a_g}$ has been shown to vary spatially and seasonally using ocean color data (Cao and Miller, 2015), consistent with photochemical fading of CDOM and seasonal riverine discharge. In oligotrophic waters, $a_g^{\prime }\left(\lambda \right)$ is overestimated if $S_{a_g}$ is higher than 0.0194 and a_g/K_d (320 nm) is close to 0.68.

Hövemoller diagrams depicting average and median monthly variations of $P_{H_2{O}_2,integ}$ are plotted in Figure 7 in 1° latitude bins. Seasonal variations in H₂O₂ photoproduction budgets in the surface mixed layer are also represented as monthly and annually integrated budgets (${\Psi }_{H_2{O}_2}$) listed in Table 5. To obtain monthly integrated budgets, $P_{H_2{O}_2,integ}$ in each pixel was multiplied by the surface area (m²) of the corresponding 1° × 1° block and the number of days in each month, and then summed. As shown in Figure 7 and Table 5, the summer months (e.g., July in the Northern Hemisphere) exhibited the highest photoproduction budget and the winter season (e.g., July in the Southern Hemisphere) exhibited the lowest photoproduction budget. The annual H₂O₂ photoproduction budget was calculated as the sum of the monthly budgets, and an annual photoproduction budget of 21.1 Tmol yr⁻¹ was obtained in this work. For comparison with Powers and Miller (2014), we also calculated monthly and annually integrated budgets (${\Psi }_{H_2{O}_2}^{\prime }$) with $P_{H_2{O}_2,integ}^{\prime }$ (data listed in Table 5). On average, the current study lowers the monthly H₂O₂ photoproduction budget estimate by 14% as compared to the modelling approach used in Powers and Miller, (2014).

FIGURE 7

FIGURE 7. Hövemoller diagrams comparing the monthly (A) average and (B) median values of $P_{H_2{O}_2,integ}$ in 1° latitude bins. White color indicates no data.

TABLE 5

TABLE 5. Estimated monthly and annually integrated H₂O₂ photoproduction budgets (${\Psi }_{H_2{O}_2}$; ${\Psi }_{H_2{O}_2}^{\prime }$). ${\Psi }_{H_2{O}_2}$ and ${\Psi }_{H_2{O}_2}^{\prime }$ were calculated with $P_{H_2{O}_2,integ}$ and $P_{H_2{O}_2,integ}^{\prime }$, respectively.

For our annual H₂O₂ photoproduction budget estimate, errors contributed by uncertainties in model inputs were evaluated following the approach used by previous global photochemical models (Fichot and Miller, 2010; Zhu and Kieber, 2020). As discussed in Section 3.2, wavelength-dependent AQYs exhibited natural variability and we expect that the errors associated with the wavelength- and temperature-dependent AQYs are within 33% and 32%, respectively. In our global model, pooled fitting parameters for the “marine-average” AQY equation and an average value of $E_a$ were used as model inputs. For wavelength-dependent AQY, the fitting parameters (m₁ and m₂) and the 95% CI are listed in Table 1; the 95% CI for m₁ and m₂ ranged from 7.30 to 7.40 and 0.0243–0.0305 m⁻¹, respectively. Using this range of model inputs for m₁ and m₂, corresponding global H₂O₂ photoproduction varied from 18.1 to 24.7 Tmol yr⁻¹ and output uncertainties contributed by the pooled m₁ and m₂ values were <17%. For $E_a$, an average ± SD of 21.8 ± 8.2 kJ mol⁻¹ was calculated from the activation energy data ($\lambda$ = 300–360 nm) reported in Kieber et al. (2014). When $E_a$ was varied from 13.6 to 30 kJ mol⁻¹, the global model resulted in H₂O₂ photoproduction values ranging between 20.8 and 21.6 Tmol yr⁻¹, demonstrating that uncertainties contributed by using an average $E_a$ value were < 2%. We also performed a sensitivity analysis by adding 30% random noise to ${\Phi }_{H_2{O}_2,T}$, $E_{{clear,0}^{-}}$, CF, $a_g$, $K_d$ and $E_a$. Readers are referred to Zhu and Kieber (2020) for a detailed description of the statistical approach used to evaluate the dependence of our global model outputs on errors associated with input parameters. Similar to Zhu and Kieber (2020) for carbonyl compounds, we estimate about a 37% uncertainty in predicted mixed-layer H₂O₂ photoproduction budgets. This translates to an uncertainty of 7.5 Tmol yr⁻¹ in H₂O₂ photoproduction.

4 Discussion

4.1 Comparison of photochemical and biological H₂O₂ production

Results from this study can be used to compare photochemical and biological production of H₂O₂ in the marine environment. Palenik and Morel (1988) found evidence of a biological H₂O₂ source between 40 and 60 m in the Sargasso Sea, which they estimated to be on the order of 1–3 nM h⁻¹. Subsequent work in the Gulf of Alaska near Kodiak Island reported biological H₂O₂ production rates ranging from 1.1—1.9 nM h⁻¹ in samples collected at 10 m based on H₂O₂-spiking experiments and modeling results (Vermilyea et al., 2010). Using a similar approach, Roe et al. (2016) found average biological H₂O₂ production rates at Station ALOHA to be 1.2 ± 0.9 nM h⁻¹ at 25 m, 0.5 ± 0.2 nM h⁻¹ at 75 m, and 0.1 ± 0.1 nM h⁻¹ at the deep chlorophyll maximum, approximately 105–130 m. Intriguingly, biological O₂⁻ production rates were variable and generally lower in the same samples, amounting to 0.25 ± 0.14 nM h⁻¹ at 25 m and 0.11 ± 0.25 at 75 m corresponding to the deep chlorophyll maximum (Roe et al., 2016). This implies that biologically produced H₂O₂ cannot result exclusively from extracellular dismutation. Recent work evaluated H₂O₂ production during incubations of 12 strains of representative marine bacteria and found a large range in rates from below detection to 300 amol cell⁻¹ h⁻¹ (Bond et al., 2020). In comparing these rates to reported O₂⁻ production rates from the same bacteria, some species had rates of O₂⁻ and H₂O₂ that were comparable in magnitude, while others were different by orders of magnitude, again suggesting that the mechanisms whereby these two ROS are released to seawater may not be quantitively linked in biological systems and thus estimates that H₂O₂ is contributed to open ocean surface seawater at a rate of ∼0.15 nM h⁻¹ may be most reasonable (Bond et al., 2020).

If we consider a biological H₂O₂ production from 0.1—1 nM h⁻¹ and a surface area of the ocean of 3.62 × 10¹⁴ m² (Charette and Smith, 2010), a very crude calculation of the biological production of H₂O₂ in the top 100 m yields an annual production of ∼ 32–320 Tmol yr⁻¹. The lower end of this range is similar to our modeled H₂O₂ photoproduction estimate. However, this crudely calculated biological H₂O₂ production assumes a constant production rate determined from a limited number of ecological settings and species, so consequently, may vary significantly if evaluated for another time or another place. Global production estimates would likely decrease significantly if temperature dependence is considered, given that temperature typically decreases rates of biological activities (Crichton et al., 2021; Sarmento et al., 2010), e.g., by a factor of ∼2 for every 10°C decrease in temperature (Packard et al., 1975). More work is needed to determine the effects of sunlight, temperature, and other stressors on the production of both O₂⁻ and H₂O₂ by marine microorganisms to refine this estimate.

4.2 Superoxide production budget estimate

To calculate the globally integrated O₂⁻ production budget we first estimated the O₂⁻:H₂O₂ production ratio based on a survey of ratios described in the literature. Powers et al. (2015) reported a ratio of 10.1(±5.4):1, determined from pooled photoproduction rates measured in marine surface water samples collected in the Gulf of Alaska. Interestingly, direct measurements of O₂⁻ and H₂O₂ at discrete time points over prolonged irradiation of the same samples yielded ratios with an even wider range from 2.4 to 10. The reader is referred to Table 6 for a summary of O₂⁻ and H₂O₂ photoproduction ratios in the SAB and the Gulf of Alaska (Powers et al., 2015; Powers and Miller, 2016). While those ratios do not represent initial rates required for photochemical rate modeling, much work remains to refine the quantitative relationship between photoproduced O₂⁻ and H₂O₂ in marine systems. Even though O₂⁻ photoproduction can be measured at pM s⁻¹ levels, determining initial rates by linear regression of [O₂⁻] versus time before steady-state concentrations are reached will likely underestimate true production since O₂⁻ decay is significant at all points in the time series. On the other hand, assumptions made to calculate O₂⁻ photoproduction from [O₂⁻] steady-state values and measured dark decay rates may not accurately reflect the full suite of O₂⁻ reactions during solar radiation. Photoproduction rates of one-electron reductants (i.e., precursors for photoproduced O₂⁻) suggest that O₂⁻ to H₂O₂ photoproduction ratios could range from 6 to 16 (Zhang et al., 2016; Le Roux et al., 2021), though open-ocean samples have not been tested. Le Roux et al. (2021) found ratios between one-electron reductants and O₂⁻ ranging from ∼1.5 to 2 for standard reference materials. This was attributed to an inability to accurately determine O₂⁻ photoproduction rates. Recent work to model (Ma et al., 2020) and measure (Le Roux et al. in review; personal communication) light-dependent O₂⁻ decay has revealed that O₂⁻ estimates using steady-state concentrations and dark decay are in error, but again these experiments have not yet been conducted in seawater. Finally, the temperature dependence for O₂⁻ photoproduction is not known and has not been evaluated. Consequently, for our model assessment, we selected a O₂⁻:H₂O₂ ratio of 10:1 as a reasonable value based on our current knowledge, allowing a back-of-the-envelope calculation of global O₂⁻ photochemical production on the order of ∼211 Tmol yr⁻¹.

TABLE 6

TABLE 6. Reported O₂⁻ and H₂O₂ photoproduction ratios and additional ancillary information.

Direct measurements of particle-associated (i.e., biological) O₂⁻ production have been reported in several field studies. Rose et al. (2008) measured [O₂⁻] in dark (30 min removed from sunlight) seawater downstream from particles collected on a 0.2 µm filter in the eastern equatorial Pacific, and reported O₂⁻ production of <1.1 nM h⁻¹ at most depths but >1.8 nM h⁻¹ at depths between 15 and 100 m. The maximum O₂⁻ production rate reported in Rose et al. (2008) was 4.4 nM h⁻¹ at 30 m near the chlorophyll maximum at one station. In Great Barrier Reef lagoon waters, net, and presumably biological, O₂⁻ production rates were 18 ± 11 nM h⁻¹ outside of a Trichnodesnium bloom and 25 ±14 nM h⁻¹ inside the bloom (Rose et al., 2010). Similarly, net O₂⁻ production rates in samples from the Gulf of Alaska were <1– ∼20 nM h⁻¹ (Hansard et al., 2010) and were typically higher in the surface (∼5–25 nM h⁻¹) than at 50 m (<1–7 nM h⁻¹). Thus, as an alternative estimate to that reported by Sutherland et al. (2020), we applied the same approach used above to estimate biological H₂O₂ production rates (See Section 4.1) and calculate biological O₂⁻ production in the top 100 m. Assuming that biological O₂⁻ production may be 1–20 nM h⁻¹ and the surface area of the ocean is 3.62 × 10¹⁴ m² (Charette and Smith, 2010), annual O₂⁻ biological production could be on the order of 320–6,300 Tmol yr⁻¹. The lower end of the estimated range is of the same magnitude as our photochemical O₂⁻ flux estimate of ∼211 Tmol yr⁻¹ in the surface mixed layer. However, overestimation is expected regarding these biological O₂⁻ production estimates since, as in the H₂O₂ calculations, the temperature dependence of these rates was not quantified and thus were not considered.

4.3 Conclusion and implications

In this work, an annual H₂O₂ photoproduction budget of 21.1 ± 7.5 Tmol yr⁻¹ was estimated using a global-scale photochemistry model driven by remotely sensed ocean data. This annual budget is significantly higher than global estimates for other marine photoproducts: carbon monoxide (2.5–4.2 Tmol yr⁻¹) (Zafiriou et al., 2003; Stubbins et al., 2006; Fichot and Miller, 2010), and the low-molecular-weight carbonyl compounds formaldehyde (4.1 Tmol yr⁻¹), acetaldehyde (2.0 Tmol yr⁻¹), glyoxal (0.35 Tmol yr⁻¹) and methylglyoxal (0.11 Tmol yr⁻¹) (Zhu and Kieber, 2020). It is often assumed that CO₂ is the dominant carbon-based photoproduct of CDOM photochemistry but rates supporting this contention were primarily obtained from inshore waters (Miller and Zepp, 1995; Mopper and Kieber, 2002). Direct AQY measurements for open ocean seawater, however, are extremely limited. Previous estimates based on CO photochemical fluxes and a CO₂:CO ratio of 15 in riverine and coastal waters (Miller and Zepp, 1995), predict an annual CO₂ budget of 37.5–63 Tmol yr⁻¹. Our global modeled production for H₂O₂ seems to be of a similar magnitude as that for CO₂, suggesting that H₂O₂ is also a dominant photoproduct of marine photochemistry.

H₂O₂ photoproduction might serve as a better proxy than CO for estimating CO₂ photoproduction. Large variabilities were observed in CO₂:CO production ratios, greatly increasing the uncertainty of previous global estimates for CO₂ photoproduction that rely on this ratio (White et al., 2010; Reader and Miller, 2012). Powers and Miller (2015a) demonstrated that H₂O₂ and CO₂ photoproduction exhibited a well-defined relationship in a variety of marine waters and reported an average CO₂:H₂O₂ molar photoproduction ratio of 6.6 ± 1.8. Using an annual H₂O₂ photoproduction budget of 21.1 Tmol yr⁻¹ and a CO₂: H₂O₂ molar ratio of 6.6, the estimated annual CO₂ photoproduction budget is 139 Tmol yr⁻¹, which is ∼2–4 times higher than the CO₂ budget estimated with CO fluxes and a CO₂:CO ratio of 15. To further constrain the potential use of H₂O₂ as a proxy to estimate CO₂ photoproduction, much more work is needed to define the variability in CO₂: H₂O₂ ratios, particularly in blue water.

Our study provides constraints for O₂⁻ photoproduction in oceanic waters. Assuming a O₂⁻: H₂O₂ production ratio of 10:1, gives an annual O₂⁻ photoproduction budget of ∼211 Tmol yr⁻¹. Despite challenges and potentially large uncertainties with this estimate, attempts to quantify O₂⁻ photoproduction on a global scale are of great interest. The exact relationship between photochemical and biological production of O₂⁻ in marine environments may have important implications for oceanic redox chemistry and marine ecology. Sutherland et al. (2020), using a compilation of laboratory derived O₂⁻ production rates from marine microorganisms, estimated dark O₂⁻ production on the order of 3–4 Pmol yr⁻¹ in the top 200 m, almost 20 times higher than our photochemical estimate in the surface mixed layer. It should be noted again that temperature and various other environmental stress factors (e.g., UV light) could significantly influence biological O₂⁻ production rates and these have yet to be evaluated. Consequently, it may be premature to extrapolate laboratory O₂⁻ dark production rates to large geographic scales.

In this work, we presented global-scale estimates of depth-resolved H₂O₂ photoproduction in sunlit surface seawater using remotely sensed ocean color. With slight modifications, the model employed in this study can be used to investigate ROS and other photo-redox reactions having defined AQY and temperature dependences at increasingly finer spatial and temporal resolutions. Our modeling approach provides new potential for examination of varying redox conditions and relationships to other important elemental cycles (e.g., Fe, Cu, Mn, N, C) in the surface oceans on regional scales and in relation to local transient oceanographic events (e.g., upwelling, rain, eddy evolution).

Data availability statement

The original contributions presented in the study are included in the article/Supplementary Material, further inquiries can be directed to the corresponding authors.

Author contributions

WM, LP, DK, and YZ designed the work. LP and WM conducted the sample collection and photochemistry experiments in the field. YZ ran the model. LP and YZ performed the data analyses. YZ, LP, DK, and WM prepared the manuscript.

Funding

Financial support was provided by the U.S. National Science Foundation through awards to the University of Georgia (OCE-1924763 and 1635618 to WM and LP) and SUNY-ESF (OCE- 1756907 to DK).

Conflict of interest

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Supplementary material

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/frsen.2022.1009398/full#supplementary-material

References

Andrews, S. S., Caron, S., and Zafiriou, O. C. (2000). Photochemical oxygen consumption in marine waters: A major sink for colored dissolved organic matter? Limnol. Oceanogr. 45, 267–277. doi:10.4319/lo.2000.45.2.0267

CrossRef Full Text | Google Scholar

Bond, R. J., Hansel, C. M., and Voelker, B. M. (2020). Heterotrophic bacteria exhibit a wide range of rates of extracellular production and decay of hydrogen peroxide. Front. Mar. Sci. 7, 00072. doi:10.3389/fmars.2020.00072

CrossRef Full Text | Google Scholar

Cao, F., Fichot, C. G., Hooker, S. B., and Miller, W. L. (2014). Improved algorithms for accurate retrieval of UV/visible diffuse attenuation coefficients in optically complex, inshore waters. Remote Sens. Environ. 144, 11–27. doi:10.1016/J.RSE.2014.01.003

CrossRef Full Text | Google Scholar

Cao, F., and Miller, W. L. (2015). A new algorithm to retrieve chromophoric dissolved organic matter (CDOM) absorption spectra in the UV from ocean color. J. Geophys. Res. Oceans 120, 496–516. doi:10.1002/2014JC010241

CrossRef Full Text | Google Scholar

Cao, F., Mishra, D. R., Schalles, J. F., and Miller, W. L. (2018). Evaluating ultraviolet (UV) based photochemistry in optically complex coastal waters using the Hyperspectral Imager for the Coastal Ocean (HICO). Estuar. Coast. Shelf Sci. 215, 199–206. doi:10.1016/J.ECSS.2018.10.013

CrossRef Full Text | Google Scholar

Charette, M. A., and Smith, W. H. F. (2010). The volume of Earth’s ocean. Oceanogr. Wash. D. C. 23, 112–114. doi:10.5670/oceanog.2010.51

CrossRef Full Text | Google Scholar

Clark, C. D., De Bruyn, W. J., Jakubowski, S. D., and Grant, S. B. (2008). Hydrogen peroxide production in marine bathing waters: Implications for fecal indicator bacteria mortality. Mar. Pollut. Bull. 56, 397–401. doi:10.1016/j.marpolbul.2007.10.017

PubMed Abstract | CrossRef Full Text | Google Scholar

Cooper, L. H. N. (1937). Oxidation-reduction potential in sea water. J. Mar. Biol. Assoc. U. K. 22, 167–176. doi:10.1017/s0025315400011929

CrossRef Full Text | Google Scholar

Cooper, W. J., and Zika, R. G. (1983). Photochemical formation of hydrogen peroxide in surface and ground waters exposed to sunlight. Science 220, 711–712. doi:10.1126/science.220.4598.711

PubMed Abstract | CrossRef Full Text | Google Scholar

Crichton, K. A., Wilson, J. D., Ridgwell, A., and Pearson, P. N. (2021). Calibration of temperature-dependent ocean microbial processes in the cGENIE.muffin (v0.9.13) Earth system model. Geosci. Model Dev. 14, 125–149. doi:10.5194/gmd-14-125-2021

CrossRef Full Text | Google Scholar

Diaz, J. M., Hansel, C. M., Voelker, B. M., Mendes, C. M., Andeer, P. F., and Zhang, T. (2013). Widespread production of extracellular superoxide by heterotrophic bacteria. Science 340, 1223–1226. doi:10.1126/science.1237331

PubMed Abstract | CrossRef Full Text | Google Scholar

Fichot, C. G., and Miller, W. L. (2010). An approach to quantify depth-resolved marine photochemical fluxes using remote sensing: Application to carbon monoxide (CO) photoproduction. Remote Sens. Environ. 114, 1363–1377. doi:10.1016/J.RSE.2010.01.019

CrossRef Full Text | Google Scholar

Fichot, C. G., Sathyendranath, S., and Miller, W. L. (2008). SeaUV and SeaUV_C: Algorithms for the retrieval of UV/Visible diffuse attenuation coefficients from ocean color. Remote Sens. Environ. 112, 1584–1602. doi:10.1016/J.RSE.2007.08.009

CrossRef Full Text | Google Scholar

Galí, M., Kieber, D. J., Romera-Castillo, C., Kinsey, J. D., Devred, E., Pérez, G. L., et al. (2016). CDOM sources and photobleaching control quantum yields for oceanic DMS photolysis. Environ. Sci. Technol. 50, 13361–13370. doi:10.1021/acs.est.6b04278

PubMed Abstract | CrossRef Full Text | Google Scholar

Goldstone, J. V., Pullin, M. J., Bertilsson, S., and Voelker, B. M. (2002). Reactions of hydroxyl radical with humic substances: Bleaching, mineralization, and production of bioavailable carbon substrates. Environ. Sci. Technol. 36, 364–372. doi:10.1021/es0109646

PubMed Abstract | CrossRef Full Text | Google Scholar

Goldstone, J. V., and Voelker, B. M. (2000). Chemistry of superoxide radical in seawater: CDOM associated sink of superoxide in coastal waters. Environ. Sci. Technol. 34, 1043–1048. doi:10.1021/es9905445

CrossRef Full Text | Google Scholar

Gueymard, C. A. (2001). Parameterized transmittance model for direct beam and circumsolar spectral irradiance. Sol. Energy 71, 325–346. doi:10.1016/S0038-092X(01)00054-8

CrossRef Full Text | Google Scholar

Haag, W. R., and Hoigne, J. (1986). Singlet oxygen in surface waters. 3. Photochemical formation and steady-state concentrations in various types of waters. Environ. Sci. Technol. 20, 341–348. doi:10.1021/es00146a005

PubMed Abstract | CrossRef Full Text | Google Scholar

Hansard, P. S., Vermilyea, A. W., and Voelker, B. M. (2010). Measurements of superoxide radical concentration and decay kinetics in the Gulf of Alaska. Deep Sea Res. Part I Oceanogr. Res. Pap. 57, 1111–1119. doi:10.1016/j.dsr.2010.05.007

CrossRef Full Text | Google Scholar

Johannessen, S. C. (2003). Calculation of UV attenuation and colored dissolved organic matter absorption spectra from measurements of ocean color. J. Geophys. Res. 108, 3301. doi:10.1029/2000JC000514

CrossRef Full Text | Google Scholar

Keeling, R. F., Körtzinger, A., and Gruber, N. (2009). Ocean deoxygenation in a warming world. Ann. Rev. Mar. Sci. 2, 199–229. doi:10.1146/annurev.marine.010908.163855

CrossRef Full Text | Google Scholar

Kieber, D. J., Miller, G. W., Neale, P. J., and Mopper, K. (2014). Wavelength and temperature-dependent apparent quantum yields for photochemical formation of hydrogen peroxide in seawater. Environ. Sci. Process. Impacts 16, 777–791. doi:10.1039/C4EM00036F

PubMed Abstract | CrossRef Full Text | Google Scholar

Kieber, D. J., Peake, B. M., Neale, P. J., and Mopper, K. (2003). “Reactive oxygen species in aquatic ecosystems,” in UV effects in aquatic ecosystems. Editors E. V. Helbling, and H. Zagarese (Cambridge, UK: Royal Society of Chemistry), 251–288.

Google Scholar

King, D. W., Cooper, W. J., Rusak, S. a., Peake, B. M., Kiddle, J. J., O’Sullivan, D. W., et al. (2007). Flow injection analysis of H₂O₂ in natural waters using acridinium ester chemiluminescence: Method development and optimization using a kinetic model. Anal. Chem. 79, 4169–4176. doi:10.1021/ac062228w

PubMed Abstract | CrossRef Full Text | Google Scholar

Le Roux, D. M., Powers, L. C., and Blough, N. V. (2021). Photoproduction rates of one-electron reductants by chromophoric dissolved organic matter via fluorescence spectroscopy: Comparison with superoxide and hydrogen peroxide rates. Environ. Sci. Technol. 55, 12095–12105. doi:10.1021/acs.est.1c04043

PubMed Abstract | CrossRef Full Text | Google Scholar

Learman, D. R., Voelker, B. M., Vazquez-Rodriguez, A. I., and Hansel, C. M. (2011). Formation of manganese oxides by bacterially generated superoxide. Nat. Geosci. 4, 95–98. doi:10.1038/ngeo1055

CrossRef Full Text | Google Scholar

Lu, W., Ridgwell, A., Thomas, E., Hardisty, D. S., Luo, G., Algeo, T. J., et al. (2018). Late inception of a resiliently oxygenated upper ocean. Science 361, 174–177. doi:10.1126/science.aar5372

PubMed Abstract | CrossRef Full Text | Google Scholar

Ma, J., Nie, J., Zhou, H., Wang, H., Lian, L., Yan, S., et al. (2020). Kinetic consideration of photochemical formation and decay of superoxide radical in dissolved organic matter solutions. Environ. Sci. Technol. 54, 3199–3208. doi:10.1021/acs.est.9b06018

PubMed Abstract | CrossRef Full Text | Google Scholar

Miller, C. J., Rose, A. L., and Waite, T. D. (2009). Impact of natural organic matter on H₂O₂-mediated oxidation of Fe(II) in a simulated freshwater system. Geochim. Cosmochim. Acta 73, 2758–2768. doi:10.1016/j.gca.2009.02.027

CrossRef Full Text | Google Scholar

Miller, W. L., and Kester, D. R. (1988). Hydrogen peroxide measurement in seawater by (p-Hydroxyphenyl)acetic acid dimerization. Anal. Chem. 60, 2711–2715. doi:10.1021/ac00175a014

CrossRef Full Text | Google Scholar

Miller, W. L., and Zepp, R. G. (1995). Photochemical production of dissolved inorganic carbon from terrestrial organic matter: Significance to the oceanic organic carbon cycle. Geophys. Res. Lett. 22, 417–420. doi:10.1029/94GL03344

CrossRef Full Text | Google Scholar

Moffett, J. W., and Zika, R. G. (1987). Reaction kinetics of hydrogen peroxide with copper and iron in seawater. Environ. Sci. Technol. 21, 804–810. doi:10.1021/es00162a012

PubMed Abstract | CrossRef Full Text | Google Scholar

Mopper, K., Kieber, D. J., and Stubbins, A. (2015). “Marine photochemistry of organic matter: Processes and impacts,” in Biogeochemistry of marine dissolved organic matter. Editors D. A. Hansell, and C. A. Carlson. 2nd Ed. (San Diego, CA: Academic Press), 389–450. doi:10.1016/B978-0-12-405940-5.00008-X

CrossRef Full Text | Google Scholar

Mopper, K. W., and Kieber, D. J. (2002). “Chapter 9 – photochemistry and the cycling of carbon, sulfur, nitrogen and phosphorus,” in Biogeochemistry of marine dissolved organic matter. Editors D. A. Hansell, and C. A. Carlson (San Diego, CA: Academic Press), 455–507. doi:10.1016/B978-012323841-2/50011-7

CrossRef Full Text | Google Scholar

Mopper, K., and Zhou, X. (1990). Hydroxyl radical photoproduction in the sea and its potential impact on marine processes. Science 250, 661–664. doi:10.1126/science.250.4981.661

PubMed Abstract | CrossRef Full Text | Google Scholar

Morel, F. M. M., and Price, N. M. (2003). The biogeochemical cycles of trace metals in the oceans. Science 300, 944–947. doi:10.1126/science.1083545

PubMed Abstract | CrossRef Full Text | Google Scholar

Morris, J. J., Johnson, Z. I., Szul, M. J., Keller, M., and Zinser, E. R. (2011). Dependence of the cyanobacterium prochlorococcus on hydrogen peroxide scavenging microbes for growth at the ocean’s surface. PLoS One 6, e16805. doi:10.1371/journal.pone.0016805

PubMed Abstract | CrossRef Full Text | Google Scholar

Ossola, R., Jönsson, O. M., Moor, K., and McNeill, K. (2021). Singlet oxygen quantum yields in environmental waters. Chem. Rev. 121, 4100–4146. doi:10.1021/acs.chemrev.0c00781

PubMed Abstract | CrossRef Full Text | Google Scholar

O’Sullivan, D. W., Neale, P. J., Coffin, R. B., Boyd, T. J., and Osburn, C. L. (2005). Photochemical production of hydrogen peroxide and methylhydroperoxide in coastal waters. Mar. Chem. 97, 14–33. doi:10.1016/j.marchem.2005.04.003

CrossRef Full Text | Google Scholar

Packard, T. T., Devol, A. H., and King, F. D. (1975). The effect of temperature on the respiratory electron transport system in marine plankton. Deep Sea Res. Oceanogr. Abstr. 22, 237–249. doi:10.1016/0011-7471(75)90029-7

CrossRef Full Text | Google Scholar

Palenik, B., and Morel, F. M. M. (1988). Dark production of H₂O₂ in the Sargasso Sea. Limnol. Oceanogr. 33, 1606–1611. doi:10.4319/lo.1988.33.6part2.1606

CrossRef Full Text | Google Scholar

Pawlowicz, R. (2020). M_Map: A mapping package for MATLAB. version 1.4m, [Computer software]. Available at: www.eoas.ubc.ca/∼rich/map.html

Google Scholar

Petasne, R. G., and Zika, R. G. (1987). Fate of superoxide in coastal sea water. Nature 325, 516–518. doi:10.1038/325516a0

CrossRef Full Text | Google Scholar

Powers, L. C., Babcock-Adams, L. C., Enright, J. K., and Miller, W. L. (2015). Probing the photochemical reactivity of deep ocean refractory carbon (DORC): Lessons from hydrogen peroxide and superoxide kinetics. Mar. Chem. 177, 306–317. doi:10.1016/j.marchem.2015.06.005

CrossRef Full Text | Google Scholar

Powers, L. C., and Miller, W. L. (2016). Apparent quantum efficiency spectra for superoxide photoproduction and its formation of hydrogen peroxide in natural waters. Front. Mar. Sci. 3, 00235. doi:10.3389/fmars.2016.00235

CrossRef Full Text | Google Scholar

Powers, L. C., and Miller, W. L. (2014). Blending remote sensing data products to estimate photochemical production of hydrogen peroxide and superoxide in the surface ocean. Environ. Sci. Process. Impacts 16, 792–806. doi:10.1039/C3EM00617D

PubMed Abstract | CrossRef Full Text | Google Scholar

Powers, L. C., and Miller, W. L. (2015a). Hydrogen peroxide and superoxide photoproduction in diverse marine waters: A simple proxy for estimating direct CO₂ photochemical fluxes. Geophys. Res. Lett. 42, 7696–7704. doi:10.1002/2015GL065669

CrossRef Full Text | Google Scholar

Powers, L. C., and Miller, W. L. (2015b). Photochemical production of CO and CO₂ in the Northern Gulf of Mexico: Estimates and challenges for quantifying the impact of photochemistry on carbon cycles. Mar. Chem. 171, 21–35. doi:10.1016/j.marchem.2015.02.004

CrossRef Full Text | Google Scholar

Reader, H. E., and Miller, W. L. (2012). Variability of carbon monoxide and carbon dioxide apparent quantum yield spectra in three coastal estuaries of the South Atlantic Bight. Biogeosciences 9, 4279–4294. doi:10.5194/bg-9-4279-2012

CrossRef Full Text | Google Scholar

Roe, K. L., Schneider, R. J., Hansel, C. M., and Voelker, B. M. (2016). Measurement of dark, particle-generated superoxide and hydrogen peroxide production and decay in the subtropical and temperate North Pacific Ocean. Deep Sea Res. Part I Oceanogr. Res. Pap. 107, 59–69. doi:10.1016/j.dsr.2015.10.012

CrossRef Full Text | Google Scholar

Rose, A. L., Godrant, A., Godrant, A., Furnas, M., and Waite, T. D. (2010). Dynamics of nonphotochemical superoxide production in the Great Barrier Reef lagoon. Limnol. Oceanogr. 55, 1521–1536. doi:10.4319/lo.2010.55.4.1521

CrossRef Full Text | Google Scholar

Rose, A. L. (2016). The influence of reactive oxygen species on local redox conditions in oxygenated natural waters. Front. Earth Sci. 4, 1–11. doi:10.3389/feart.2016.00096

CrossRef Full Text | Google Scholar

Rose, A. L., and Waite, T. D. (2006). Role of superoxide in the photochemical reduction of iron in seawater. Geochim. Cosmochim. Acta 70, 3869–3882. doi:10.1016/j.gca.2006.06.008

CrossRef Full Text | Google Scholar

Rose, A. L., Webb, E. A., Waite, T. D., and Moffett, J. W. (2008). Measurement and implications of nonphotochemically generated superoxide in the equatorial Pacific Ocean. Environ. Sci. Technol. 42, 2387–2393. doi:10.1021/es7024609

PubMed Abstract | CrossRef Full Text | Google Scholar

Sarmento, H., Montoya, J. M., Vázquez-Domínguez, E., Vaqué, D., and Gasol, J. M. (2010). Warming effects on marine microbial food web processes: How far can we go when it comes to predictions? Philos. Trans. R. Soc. B Biol. Sci. 365, 2137–2149. doi:10.1098/rstb.2010.0045

PubMed Abstract | CrossRef Full Text | Google Scholar

Sawyer, D. T. (1991). Oxygen chemistry. Oxford, United Kingdom: Oxford University Press.

Google Scholar

Schmidtko, S., Johnson, G. C., and Lyman, J. M. (2013). Mimoc: A global monthly isopycnal upper-Ocean Climatology with mixed layers. J. Geophys. Res. Oceans 118, 1658–1672. doi:10.1002/jgrc.20122

CrossRef Full Text | Google Scholar

Scully, N. M., Cooper, W. J., and Tranvik, L. J. (2003). Photochemical effects on microbial activity in natural waters: The interaction of reactive oxygen species and dissolved organic matter. FEMS Microbiol. Ecol. 46, 353–357. doi:10.1016/S0168-6496(03)00198-3

PubMed Abstract | CrossRef Full Text | Google Scholar

Sharpless, C. M., and Blough, N. V. (2014). The importance of charge-transfer interactions in determining chromophoric dissolved organic matter (CDOM) optical and photochemical properties. Environ. Sci. Process. Impacts 16, 654–671. doi:10.1039/C3EM00573A

PubMed Abstract | CrossRef Full Text | Google Scholar

Stubbins, A., Uher, G., Law, C. S., Mopper, K., Robinson, C., and Upstill-Goddard, R. C. (2006). Open-ocean carbon monoxide photoproduction. Deep Sea Res. Part II Top. Stud. Oceanogr. 53, 1695–1705. doi:10.1016/j.dsr2.2006.05.011

CrossRef Full Text | Google Scholar

Sutherland, K. M., Grabb, K. C., Karolewski, J. S., Taenzer, L., Hansel, C. M., and Wankel, S. D. (2021). The redox fate of hydrogen peroxide in the marine water column. Limnol. Oceanogr. 66, 3828–3841. doi:10.1002/lno.11922

CrossRef Full Text | Google Scholar

Sutherland, K. M., Wankel, S. D., and Hansel, C. M. (2020). Dark biological superoxide production as a significant flux and sink of marine dissolved oxygen. Proc. Natl. Acad. Sci. U. S. A. 117, 3433–3439. doi:10.1073/pnas.1912313117

PubMed Abstract | CrossRef Full Text | Google Scholar

Tanskanen, A., Krotkov, N. A., Herman, J. R., and Arola, A. (2006). Surface ultraviolet irradiance from OMI. IEEE Trans. Geosci. Remote Sens. 44, 1267–1271. doi:10.1109/TGRS.2005.862203

CrossRef Full Text | Google Scholar

Vermilyea, A. W., Paul Hansard, S., and Voelker, B. M. (2010). Dark production of hydrogen peroxide in the Gulf of Alaska. Limnol. Oceanogr. 55, 580–588. doi:10.4319/lo.2010.55.2.0580

CrossRef Full Text | Google Scholar

Voelker, B. M., and Sulzberger, B. (1996). Effects of fulvic acid on Fe(II) oxidation by hydrogen peroxide. Environ. Sci. Technol. 30, 1106–1114. doi:10.1021/es9502132

CrossRef Full Text | Google Scholar

White, E. M., Kieber, D. J., Sherrard, J., Miller, W. L., and Mopper, K. (2010). Carbon dioxide and carbon monoxide photoproduction quantum yields in the Delaware Estuary. Mar. Chem. 118, 11–21. doi:10.1016/j.marchem.2009.10.001

CrossRef Full Text | Google Scholar

White, E. M., Vaughan, P. P., and Zepp, R. G. (2003). Role of the photo-Fenton reaction in the production of hydroxyl radicals and photobleaching of colored dissolved organic matter in a coastal river of the southeastern United States. Aquatic Sci. - Res. Across Boundaries 65, 402–414. doi:10.1007/s00027-003-0675-4

CrossRef Full Text | Google Scholar

Wuttig, K., Heller, M. I., and Croot, P. L. (2013). Pathways of superoxide (O₂^–) decay in the eastern tropical north atlantic. Environ. Sci. Technol. 47, 10249–10256. doi:10.1021/es401658t

PubMed Abstract | CrossRef Full Text | Google Scholar

Yocis, B. H., Kieber, D. J., and Mopper, K. (2000). Photochemical production of hydrogen peroxide in Antarctic Waters. Deep Sea Res. Part I Oceanogr. Res. Pap. 47, 1077–1099. doi:10.1016/S0967-0637(99)00095-3

CrossRef Full Text | Google Scholar

Zafiriou, O. C., Andrews, S. S., and Wang, W. (2003). Concordant estimates of oceanic carbon monoxide source and sink processes in the Pacific yield a balanced global “blue-water” CO budget. Glob. Biogeochem. Cycles 17, 1015. doi:10.1029/2001GB001638

CrossRef Full Text | Google Scholar

Zepp, R. G., Callaghan, T. V., and Erickson, D. J. (2003). Interactive effects of ozone depletion and climate change on biogeochemical cycles. Photochem. Photobiol. Sci. 2, 51–61. doi:10.1039/B211154N

PubMed Abstract | CrossRef Full Text | Google Scholar

Zepp, R. G., Faust, B. C., and Hoigne, J. (1992). Hydroxyl radical formation in aqueous reactions (pH 3-8) of iron(II) with hydrogen peroxide: The photo-fenton reaction. Environ. Sci. Technol. 26, 313–319. doi:10.1021/es00026a011

CrossRef Full Text | Google Scholar

Zhang, F., Xiao, S., Kendall, B., Romaniello, S. J., Cui, H., Meyer, M., et al. (2018). Extensive marine anoxia during the terminal Ediacaran Period. Sci. Adv. 4, eaan8983. doi:10.1126/sciadv.aan8983

PubMed Abstract | CrossRef Full Text | Google Scholar

Zhang, T., Diaz, J. M., Brighi, C., Parsons, R. J., McNally, S., Apprill, A., et al. (2016). Dark production of extracellular superoxide by the coral porites astreoides and representative symbionts. Front. Mar. Sci. 3, 00232. doi:10.3389/fmars.2016.00232

CrossRef Full Text | Google Scholar

Zhang, Y., and Blough, N. V. (2016). Photoproduction of one-electron reducing intermediates by chromophoric dissolved organic matter (CDOM): Relation to O₂^- and H₂O₂ photoproduction and CDOM photooxidation. Environ. Sci. Technol. 50, 11008–11015. doi:10.1021/acs.est.6b02919

PubMed Abstract | CrossRef Full Text | Google Scholar

Zhu, Y., and Kieber, D. J. (2019). Concentrations and photochemistry of acetaldehyde, glyoxal and methylglyoxal in the Northwest Atlantic Ocean. Environ. Sci. Technol. 53, 9512–9521. doi:10.1021/acs.est.9b01631

PubMed Abstract | CrossRef Full Text | Google Scholar

Zhu, Y., and Kieber, D. J. (2020). Global model for depth-dependent carbonyl photochemical production rates in seawater. Glob. Biogeochem. Cycles 34, 1–20. doi:10.1029/2019GB006431

CrossRef Full Text | Google Scholar

Zhu, Y., and Kieber, D. J. (2018). Wavelength and temperature-dependent apparent quantum yields for photochemical production of carbonyl compounds in the North Pacific Ocean. Environ. Sci. Technol. 52, 1929–1939. doi:10.1021/acs.est.7b05462

PubMed Abstract | CrossRef Full Text | Google Scholar