Fresh fronds of U. prolifera and Sargassum were collected during the bloom stage from the coast of Qingdao, China in June 2021. Immediately after sampling, they were taken to the laboratory, thoroughly rinsed with autoclave-sterilized seawater (115°C, 30 min), air-dried, and refrigerated at -20°C for further treatments. The fronds (10 g) were cut into small pieces using a ceramic knife and ground to a finer consistency in an agate mortar. The ground fronds were added to 2 L of artificial seawater (salinity 36.03, ASTM D1141-98, 2021) in a glass beaker and ultrasonic-pulverized for 10 min. The suspension was prefiltered through a 6.5-μm bolting silk, followed by filtration sequentially through glass microfiber GF/F filters (Whatman) and 0.20 μm Nylon membranes (Millipore) to remove particles including microorganisms. The final absorbances of the filtrates were below 1.0 (cell pathlength: 1 cm) at 280 nm so that Beer-Lambert Law was obeyed. The corresponding concentration of dissolved organic carbon (DOC) was 959.6 μmol L^-1 for the U. prolifera filtrate and 3537.8 μmol L^-1 for the Sargassum filtrate.

The filtrate was then bubbled with a mixture of nitrogen (N₂) and oxygen (O₂) with a mole ratio of 79:21 to reduce the background content of CO₂, CO, and CH₄. For CO and CH₄ samples, the filtrate’s pH increased during bubbling and was adjusted to its original value (7.36) using 0.10 mol L^-1 hydrochloric acid; whereas for CO₂ samples, the samples’ pH was preadjusted to 4 with 1.0 mol L^-1 HCl before bubbling and was adjusted back to the original value using 0.10 mol L^-1 NaOH after bubbling. The filtrate was then siphoned into 95.0-mL cylindrical quartz tubes (length: 25.0 cm; i.d.: 2.2 cm). The tubes were sealed without headspace using ground glass stoppers following profuse overflow.

Before use, the GF/F filters were pre-combusted at 450°C for 5 h and Nylon membranes were thoroughly rinsed with Mill-Q water; all glassware was thoroughly rinsed with Milli-Q water, air-dried, and then combusted at 450°C for 5 h.

Irradiations were performed using a solar simulator (Q-SUN Xe-1, Q-Lab Corporation, USA) equipped with a 1800-W xenon lamp. A special UV glass filter was installed to remove UV radiation with wavelengths< 290 nm. The sample-filled quartz tubes were horizontally immersed (~2 mm below the water surface) in a temperature-controlled water bath (20.0 ± 0.5°C) located immediately beneath the exposure chamber of the solar simulator. The samples were irradiated under full spectrum for 10 min to determine the photoproduction rate of CO and for 24 h to determine the photoproduction rates of CO₂ and CH₄. All irradiations were accompanied by dark controls. Samples were incubated and analyzed in triplicate.

The photon fluxes of the solar simulator at the upper surface of the irradiation cells were measured using an OL-756 spectroradiometer fitted with a 2-inch OL IS-270 integrating sphere and calibrated using an OL 752-10E irradiance standard. Figure 2 shows a comparison of the solar simulator’s photon flux spectrum with those of sunlight recorded hourly on June 30, 2023, in Qingdao, China (36.369°N, 120.690°E). The solar simulator’s photon flux integrated over the UVB region (280–320 nm) was 0.57 times that of sunlight measured at 11:30, 0.89 times over the UVA region (320–400 nm), and 0.59 times over the VIS region (400–600 nm). Summing these hourly solar photon fluxes yields the daily photon flux in Qingdao on June 30, 2023. The 24-h simulated irradiation for the full spectrum (280–600 nm) corresponds to 2.1 days of the solar irradiation on that specific date (1.88 days for the UVB band, 2.81 days for the UVA band and 1.85 days for the VIS band).

CH₄ and CO were measured using a static headspace method as described by Zhang et al. (2020). Briefly, water samples were transferred to a 50-mL glass syringe fitted with three-way valves (glass fiber-reinforced polypropylene). The syringe was rinsed with sample water before the final drawing. Then 5 mL N₂ were introduced into the syringe to obtain a 1:6 gas-to-water ratio. The syringe was vigorously shaken for 6 min and the equilibrated headspace gas was injected into a Peak Performer 1 FID gas chromatograph (1-mL sample loop; Peak laboratories, USA) for quantification of CH₄ and CO. The analyzer was equipped with a methanizer to convert CO to CH₄ and standardized by frequent injections of a gaseous standard containing 5.00 ppmv CH₄ and 4.43 ppmv CO in pure N₂ (National Institute of Metrology, China). In keeping with the sample’s 100% relative humidity, the dry standard gas was saturated with water vapor before injection. To estimate the analytical blank, a water sample was repeatedly extracted with pure N₂ until the CH₄ and CO signals diminished to stable levels. For CH₄, ten sequential analyses of the extracted sample arrived at a mean blank of 0.007 nmol kg⁻¹ with a standard deviation of 0.002 nmol kg⁻¹. The lower detection limit, defined as three times the standard deviation, was thus 0.006 nmol kg⁻¹. For CO, the blank was 0.003 nmol kg^-1 and the lower detection limit was 0.007 nmol kg^-1. The analytical reproducibility was determined to be ± 4% (n = 10) for CH₄ at a concentration of ~5 nmol kg⁻¹ and ± 6% (n = 10) for CO at a concentration of ~4 nmol kg^-1.

CO₂ (in the form of dissolved inorganic carbon, DIC) was measured using an infrared CO₂ detector-based AS-C3 DIC Analyzer (Apollo SciTech Inc., USA) calibrated against the Certificated Reference Materials from Andrew G. Dickson’s lab at the Scripps Institution of Oceanography, with a precision of ± 2 μmol kg^-1 (Chen et al., 2022). The amounts of photochemically produced CO₂, CO, and CH₄ were calculated as the differences between the irradiated samples and the parallel dark controls.

Absorbance spectra (250–600 nm, 1-nm intervals) of the filtered water samples were recorded at room temperature using a UV-visible spectrophotometer (Cary 100, Agilent, USA). The samples were placed in 1-cm quartz cuvettes and referenced to Milli-Q water. Absorbance was baseline-corrected by subtracting the average absorbance over an interval of 5 nm around 685 nm (Babin et al., 2003). The Napierian absorption coefficient, a _CDOM(λ) (m⁻¹), where λ is wavelength in nanometers, was calculated as 2.303 times the absorbance divided by the cuvette’s pathlength in meters (0.01 m). The lower detection limit for the absorption coefficient measurement, defined as three times the standard deviation of five replicate analyses of pure water, was determined to be 0.02 ± 0.01 m^-1 over 280–600 nm.

Spectral slope coefficients over 275–295 nm (S _275-295) of CDOM absorption spectra were calculated using linear regression of the log-transformed absorption spectra, following the method of Helms et al. (2008). S _275–295 has been used as an indicator of the mean molecular weight of CDOM, with higher S _275–295 values associated with lower molecular weights (Helms et al., 2008).

DOC was quantified using a TOC-L Analyzer (Shimadzu, Japan) equipped with an ASI-L autosampler. All samples were pre-acidified with H₃PO₄ to pH = 2. The instrument was calibrated using potassium hydrogen phthalate standard solutions and checked every six sample runs against the reference deep seawater (DOC: 41–44 μmol L^-1) provided by the Hansell laboratory at the University of Miami. The relative standard deviation of replicate measurements of the reference deep seawater was approximately 2%. Instrumental blanks were determined using Milli-Q water and deducted from the samples’ results.

An Orion Versa Star Pro benchtop meter (Thermo Scientific) fitted with a Ross Ultra pH electrode (Orion 8157 BNUMD) was used to determine pH; the system was standardized with three NIST buffers at pH 4.01, 7.00 and 10.01.

The photon flux absorbed by CDOM at wavelength λ (nm), Q _CDOM(λ) (mol photons s^-1 nm^-1), was calculated according to Hu et al. (2002):

In Equation 1, Q ₀ (mol photons m⁻² s⁻¹ nm⁻¹) is the photon flux at the upper water surface inside the quartz cell; a _CDOM(λ)(m^-1) the geometric mean of the absorption coefficients measured before and after irradiation; a _t(λ) (m^-1) the sum of a _CDOM(λ) and the spectral absorption coefficient of pure water (Buiteveld et al., 1994; Pope and Fry, 1997); S the longitudinal cross-section of the quartz tube (0.0042 m²); L the light pathlength of the tube (0.0268 m), calculated as the squared root of the latitudinal cross-section of the tube (Osburn et al., 2001).

AQY, defined as the number of moles of a photoproduct formed per mole of photons absorbed by CDOM, is used to characterize the efficiency of a given CDOM photoreaction (Hu et al., 2002). Broadband (280–500 nm) AQYs for CO₂ (AQY_CO2 in mol CO₂ (mol photons)^-1), CO (AQY_CO in mol CO (mol photons)^-1), and CH₄ (AQY_CH4 in mol CH₄ (mol photons)^-1) were calculated using Equation 2.

Since the concentrations of CDOM chromophores are unknown, apparent AQY for photobleaching was calculated as the loss of a _CDOM at a given wavelength (e.g., 330 nm, Osburn et al., 2009) divided by the number of moles of photons absorbed by CDOM (AQY_ble(330) in m^-1 (mole photons)^-1)):

In Equation 3, the wavelength of 330 nm was chosen to facilitate comparison with earlier studies (e.g., Osburn et al., 2009) and also because maximum photolysis rates of aquatic CDOM occur at or near this wavelength (Hu et al., 2002).

For comparison with earlier studies reporting spectrally resolved AQYs but without providing broadband AQYs, we calculated simulated solar spectrum-weighted mean AQYs ( AQY ¯ ) over the wavelength range of 280–500 nm according to Zhang et al. (2006):

In Equation 4, Q _λ denotes the spectral irradiance of the solar simulator used in this study ( Figure 2 ); AQY_λ the spectrally resolved AQY of absorbance photobleaching, CO₂, CO, or CH₄ reported previously. To assess the uncertainty of using AQY  ¯ for comparison with broadband AQYs, we calculated both the broadband AQY and AQY ¯ of CO (280–500 nm) for water samples collected from the Estuary and Gulf of St. Lawrence using the spectral CO AQYs and full-spectrum CO photoproduction rates obtained by Zhang et al. (2006). The ratios of the broadband AQY to AQY ¯ averaged 1.09 with a standard deviation of 0.29 (n = 54).

The absorption coefficients of the original (i.e., unirradiated) UP- and S-CDOM decreased quasi-exponentially with increasing wavelength over the UV-visible range (280–600 nm). A shoulder over 310–355 nm, however, superimposed the general trend of the S-CDOM spectrum, while the UP-CDOM spectrum lacked discernible shoulders ( Figures 3A, B ). The peak wavelength of the S-CDOM shoulder-converted peak was found to be ~330 nm ( Figure 3B inset), which is characteristic of mycosporine-like amino acids (Carreto et al., 2005). Based on the absorption coefficients averaged over 254–500 nm, respectively, Sargassum leached out 3.5 times more CDOM than did U. prolifera per unit wet weight of the macroalgae. The DOC-normalized absorption coefficient at 254 nm (a^* _CDOM(254)), an indicator of the aromaticity of DOM (Weishaar et al., 2003), was 0.49 L mgC^-1 m^-1 for UP-CDOM and 0.68 L mgC^-1 m^-1 for S-CDOM. The relatively higher value for S-CDOM is probably due to Sargassum containing a high content of phlorotannins, a class of polymers of phloroglucinol (1,3,5-trihydroxybenzene) synthesized by brown algae but absent in green algae (Stern et al., 1996; Powers et al., 2020). The disparities in the absorption spectral shape and a ^* _CDOM(254) between the UP- and S-CDOM suggest that different macroalgae may produce different compounds that possess distinct CDOM absorption features. However, the two CDOM pools displayed similar S _275-295 values (UP-CDOM: 0.0214 nm^-1; S-CDOM: 0.0208 nm^-1), implying that they had comparable average molecular weights.

The values of a^* _CDOM(254) from this study are lower than those reported in the surface Yellow Sea during spring and summer (2.0–2.5 L mg C^-1 m^-1, Yang et al., 2021). This difference may be attributed to sulfated polysaccharides present in the cell wall and intercellular substance of U. prolifera and Sargassum (Rushdi et al., 2020; Zhong et al., 2020). These polysaccharides contribute much less to the aromaticity than to the DOC content according to their molecular structures. It is also plausible that DOM in the surface Yellow Sea contains significant amounts of terrigenous materials, which are known to be enriched with aromatic moieties relative to marine DOM (Liang et al., 2023). The values of S _275-295 from this study are comparable to those reported for fresh, autochthonous CDOM in the northern Yellow Sea (range 0.0200–0.0211, mean 0.0205 ± 0.0007 nm^-1, Zhang et al., 2018) and the Bohai Sea (range 0.0228–0.0241 nm^-1, mean 0.0235 ± 0.0006 nm^-1, Zhang et al., 2022). In contrast, older CDOM in these two areas exhibited higher S _275-295 values (0.0260 and 0.0270 nm^-1, respectively) (Zhang et al., 2018, 2022).

After the 24-h irradiation, the mean absorption coefficient in the UVB, UVA, and VIS regimes decreased, respectively, by 37.3%, 39.2%, and 28.3% for the UP-CDOM ( Figure 3A ) and by 40.6%, 51.9%, and 22.9% for the S-CDOM ( Figure 3B ). UVA thus led to the largest reduction of the mean absorption coefficient for both CDOM pools, in line with the absorbed photon flux being strongest within this band ( Figure 3C ). The photobleaching increased S _275-295 by 86% (from 0.0214 to 0.0399 nm^-1) for UP-CDOM and by 49% (from 0.0208 to 0.0310 nm^-1) for S-CDOM, indicating decreases in the average molecular weight of CDOM. Notably, the characteristic shoulder in the S-CDOM spectrum persisted after the irradiation ( Figure 3B ) but the shoulder area, calculated as the integral of the shoulder a _CDOM(λ) over 310–355 nm ( Figure 3B inset), decreased by 21%. This decrease was less than the 53% reduction in the corresponding background area represented by the integral of the background a _CDOM(λ) (dotted lines) over the same wavelength range ( Figure 3B ), suggesting that the shoulder-specific compound is less prone to photobleaching than the rest of the S-CDOM.

The 24-h irradiation decreased the a _CDOM(330) by 44% (from 1.38 m^-1 to 0.78 m^-1) for UP-CDOM and by 48% (from 12.01 m^-1 to 6.29 m^-1) for S-CDOM. Following Equation (3), AQY_ble(330) is calculated to be 466 m^-1 (mole photons)^-1 for UP-CDOM and 1108 m^-1 (mole photons)^-1 for S-CDOM ( Table 1 ). S-CDOM was thus more susceptible to photobleaching than UP-CDOM. Note that the wavelength of 330 nm is within the absorption shoulder of S-CDOM. Since the shoulder-specific compound is relatively less susceptible to photobleaching than the background S-CDOM (see above), the difference in AQY_ble(330) between the two CDOM pools could be even larger if only the background S-CDOM were considered. The AQY_ble(330) values for UP- and S-CDOM are much higher than those reported previously for CDOM in the Saguenay river (154 m^-1 (mole photons)^-1) (Zhang and Xie, 2015), CDOM in groundwaters of the Îles-de-la-Madeleine in the Gulf of St. Lawrence, Canada (0.02–0.18 m^-1 (mol photons)^-1) (Qi et al., 2018) and CDOM in waters across the Mackenzie shelf of the western Canadian Arctic (0.080–0.140 m^-1 (mole photons)^-1) (Osburn et al., 2009) ( Table 1 ). This comparison indicates that the UP- and S-CDOM are far more sensitive to photobleaching than CDOM in natural waters on an absorbed-photons basis.