The observation site (120.44°E, 36.05°N) is located at Qingdao coastal area ( Figure 1 ), near Xiaomaidao Island, and is in a transition zone connecting the southern Yellow Sea with Jiaozho Bay. This site is dominated by semi-diurnal currents, with a water depth of about 23 m. The rainfall is highest in summer and lowest in winter, with an average annual precipitation of approximately 660 mm (http://www.qingdao.gov.cn). The investigation area is affected by the southeast monsoon and coastal waters in summer. The sea surface salinity (SSS) is the highest in winter, and the peak sea surface temperature (SST) generally occurs in August. Since 2007, U. prolifera has occurred annually in the southern Yellow Sea, and Qingdao is one of the cities receiving the macroalga.

The 15 m intelligent buoy was deployed at the observation site from May 29 to July 25 in 2024. There was an air-water equilibrator (MAPCO₂) (Friederich et al., 1995; Sutton et al., 2014) placed in the buoy shaft floating on the sea surface based on float, and a customized gas control module and standard gas ( Figure 2 ) fixed in the buoy chamber without contact with seawater for pCO₂ measurement. During monitoring, the seawater pCO₂ was measured with a frequency of 1 h or 2 h and the calibration frequency was 24 h. The atmosphere pCO₂ data which was measured at sit on May 29, 2024, was used for air-sea CO₂ flux calculation. The sea surface temperature (SST) and sea surface salinity (SSS) were determined by SBE37 (Sea-bird Scientific). The dissolved oxygen saturation (DO%) which refers to the ratio of seawater DO content to its solubility, was measured using the probe developed by the institute of Oceanographic Instrumentation. The wind speed was measured using an ultrasonic anemometer (windmaster pro, Gill Instruments) installed at a height of 10 m.

The in-situ non-dispersive infrared (NDIR) seawater pCO₂ instruments (Fietzek et al., 2014; Sutton et al., 2014; Hunt et al., 2017; Ribas-Ribas et al., 2018) are often employed on the buoy to describe the temporal variations of pCO₂ and CO₂ fluxes at the air-sea interface (Atamanchuk et al., 2015; Xue et al., 2016; Liu Q. et al., 2019). They used the gas-permeable membrane (Fietzek et al., 2014) or continuous bubbling air–water equilibrator (Sutton et al., 2014) to equilibrate the headspace air with seawater. After equilibration, the CO₂ composition of headspace air reflects that of seawater. The evaluation and comparison of commercial in-situ pCO₂ instruments demonstrated that the system employing bubble equilibration technique (Sutton et al., 2014) had high accuracy and stability (Alliance for Coastal Technologies, 2010a, b, c). Though the commercial in-situ instrument using bubble equilibration (MAPCO₂) is widely used, it is sometimes inflexible for installation, inconvenient self-maintenance and expensive. The water quality monitoring buoys have been well developed by integrating biochemical sensors, optical devices for absorption, fluorescence and scattering, and communication devices (Błażejewski et al., 2024, 2023; Agade et al., 2022; Ng et al., 2012) for field applications. Therefore, easily integrated seawater pCO₂ instruments are needed for future collaborative observations.

To flexible investigate the seawater pCO₂ variations in Qingdao coastal waters, a customized and concise gas control module combined with standard gas (National Standard Material Research Center) and the bubble equilibrator of MAPCO₂ (Friederich et al., 1995; Sutton et al., 2014) which connects gas control module with two gas pipelines were deployed on a 15m intelligent buoy in this article. The module operates in three modes: seawater measurement, atmosphere measurement and calibration modes ( Figure 2A ). During seawater pCO₂ measurement, the gas driven by pump circularly goes out from the gas pipe (1.6 mm i.d., 3.2 mm o.d.) beneath seawater surface and continuously bubbles in the equilibrator which floated on the sea surface, then the gas returns to the gas control module through the other pipe (3.2 mm i.d., 6.4 mm o.d.). The equilibrator was made of copper-nickel alloy to prevent biofouling. More details for bubble equilibrator can be found in the paper by Sutton et al. (2014). During monitoring, the control module timed measured standard gas by switching different modes. The 12V voltage for gas control module was provided by buoy’s solar power, and data is transmitted over 3G/4G networks.

The gas control module is housed in a sealed cylindrical cabin, including NDIR detector (Licor850, LI-COR Biotechnology), gas flow sensor (AWM3300, Honeywell Automation & Control, Inc.), desiccant (silica gel), CO₂ absorbent (soda lime), filter (1.0 µm), gas pump (KLVP-SB12, Kamoer Fluid Tech Shanghai Co., Ltd., China), electromagnetic three-way valves (LVM105R-6B, SMC), customized multi-channel plate, gas pipe and the control circuit ( Figure 2 ). The NDIR detector measures mole fraction of CO₂ (xCO₂) and H₂O (xH₂O) in gas path at about 51.5 °C, and has a thermal insulation cavity outside to ensure the stability and accuracy measurements. The desiccant is used to absorb vapor in the gas path. The soda lime is used to absorb CO₂ during zero gas calibration. The gas flow sensor is used to monitor the condition of system during measurement. The three-way valves were connected to the gas pipe through channels (diameter 1.5 mm) inside the multi-channel plate ( Figure 2B ). The gas control module could be operated in different modes based on the positions of three-way valves, similar to other instruments. The gas control module has the characteristics of simple structure and good integration.

Given the lack of drying methods for long-term autonomous measurements, the determined xCO₂ should be converted to dry xCO₂. Finally, the pCO₂ of seawater at 100% humidity was calculated.

The dry xCO₂ was calculated using the following equation:

where the xCO₂ ^dry (µmol mol^-1) is xCO₂ in dry air, and xCO₂ (µmol mol^-1) and xH₂O (mol mol^-1) denote the measured concentrations.

The partial pressure of the surface seawater CO₂ (pCO₂, µatm) is calculated as follows:

where x CO 2  corr dry is the corrected xCO₂, p denotes the pressure (atm) of atmosphere, and V P H 2 O   refers to the water vapor pressure at 100% humidity obtained using the in-situ temperature (t, °C) and salinity (S) (Weiss and Price, 1980).

The precision of the pCO₂ measurement system was evaluated through repeated measurements of seawater samples, and the results were expressed as standard deviation (SD) of xCO₂. The seawater samples were adjusted to various pCO₂ values using NaOH and HCl solutions. Table 1 shows the precision of the seawater pCO₂ measurement system during repeated measurements of three seawater samples in the laboratory. With xCO₂ values in the range of 390–800 µmol mol^-1, the measurement system exhibited a precision better than 1 µmol mol^-1, with the SD ranging from 0.67 µmol mol^-1 to 0.84 µmol mol^-1. Good precision here provides a prerequisite for obtaining good accuracy.

The accuracy was evaluated by comparing with an underway seawater pCO₂ instrument (GO8050, General Oceanics, INC.). The accuracy was expressed as the relative error of xCO₂ ( Table 2 ). Prior to comparison, the underway instrument was calibrated using standard gases. For comparison, equilibrator and water pump of the underway instrument were placed in a thermostatic bath containing the seawater samples. The accuracy was estimated in a xCO₂ range of 200 – 1000 µmol mol^-1 through adjustment of seawater with HCl and NaOH solutions. The average of 10 continuous measurements obtained by the GO8050 analyzer after equilibrium was considered the true value. In the xCO₂ range of 200 – 1000 µmol/mol, the error was -5.1 – 2.3 µmol mol^-1, and the relative error was within 1%.

To eliminate the thermodynamic influence of temperature, we normalized pCO₂ to the average temperature of the investigation (npCO₂) by Takahashi et al. (1993):

where npCO₂ is the normalized pCO₂, and t_m denotes the average temperature during the time series monitoring. Meanwhile, simulated pCO₂ changes solely caused by temperature could be calculated when SST is the initial temperature and t_m is the observed temperature.

Air–sea CO₂ flux (FCO₂, mmol m^-2d^-1) was estimated according to the following equation:

where k indicates the gas transfer velocity, K ₀ represents the solubility coefficient of CO₂ (Weiss, 1974), and pCO_2water and pCO_2air are the pCO₂ in the surface water and atmosphere, respectively. k was estimated using the formula proposed by Sweeney et al. (2007).

where U₁₀ (m s^-1) refers to the wind speed at 10 m height. Sc denotes the Schmidt number of CO₂ in seawater (Wanninkhof, 1992). In this paper, a positive value of FCO₂ denoted CO₂ release from water to the atmosphere.

During the buoy-based time series observation period of May 29 – July 25, 2024, U. prolifera bloom was observed at Qingdao coastal region based on satellite remote images. Its coverage increased in mid-July and then decreased in late July ( Figure 3 ). The SST increased from approximately 16°C to 22°C ( Figure 4A ) due to the increase in solar radiation. By contrast, SSS showed a decreased trend and ranged from ~31.0 PSU to ~29.2 PSU ( Figure 4A ). The sudden decrease in SSS during the observation period was possibly due to rainfall. On the days when the salinity dropped to about 29.2 PSU (July 8, July 16 and July 22), the daily rainfall at the observatory near Xiaomaidao Island was 58 mm, 39 mm and 48 mm, respectively (http://swglj.qingdao.gov.cn). During the observation period, the tidal height varied between 56 cm and 437 cm ( Figure 4B ). DO% also showed a general downward trend, varying within a range of 112% to 89% during the investigation period, and temporarily high values were observed in July ( Figure 4D ). Sea surface pCO₂ ranged from 519 µatm to 717 µatm, with an average of 615 ± 45 µatm, and the daily average sea surface pCO₂ increased by ~110 µatm at the end of observation compared with that in the beginning ( Figure 4C ). While the npCO₂ of seawater showed a slightly decreasing trend, and a ~30 µatm lower daily average npCO₂ was observed at the end of observation compared with that in the beginning ( Figure 4D ). In addition, both surface pCO₂ and npCO₂ showed a rapid decline and rise from July 12 to 21, and then the values remained stable.

The atmospheric pCO₂ was 417 µatm, which is lower than that of seawater. Thus, this observation site served as a CO₂ source for the atmosphere throughout the observation period in 2024. The wind speed ranged from 0.1 m s^-1 to 18.5 m s^-1, and the average was 3.0 m s^-1 ( Figure 4E ). The large variations in wind speed and seawater pCO₂ induced considerable changes in CO₂ flux. The influence of wind speed is particularly important. According to Equation 6, the influence of wind speed on CO₂ flux is quadratic. The CO₂ flux ranged from less than 1 mmol m^-2 d^-1 to over 100 mmol m^-2 d^-1, roughly consistent with wind speed, with an average value of 5.9 mmol m^-2 d^-1 ( Figure 4F ). Furthermore, the variations in CO₂ flux during the observation period demonstrated the importance of high-resolution monitoring in carbon flux estimation. Based on collected data, the seawater at investigation site near Xiaomaidao Island released 334 mmol CO₂ m^-2 to the atmosphere during May 29 – July 25, 2024.

The data collected in the western part of the southern Yellow Sea which lies approximately between 33°N and 37°N and west of the 50m isobath (Wang and Zhai, 2021) and Jiaozhou Bay (Li et al., 2023) during summer were compared with the investigation results of this study ( Table 3 ). It was shown that the western part of the southern Yellow Sea had relatively lower SST, seawater pCO₂, CO₂ flux, and higher SSS, while lower SSS accompanied with higher SST, seawater pCO₂ and CO₂ flux were observed in Jiaozhou Bay compared with this study. During summer the Jiaozhou Bay became a strong source of atmospheric CO₂. In addition to the higher SST, terrestrial inputs were also the reason for the high sea surface pCO₂ and CO₂ flux in Jiaozhou Bay. It was reported that except for the Dagu River, most rivers in Jiaozhou Bay had no pristine runoff and almost became channels of wastewater which was abundant in pCO₂, resulting in the high seawater pCO₂ values observed in Jiaozhou Bay (Gao et al., 2008; Li et al., 2023). Though there was no freshwater input from rivers near Xiaomaidao Island, the Qingdao coast area was significantly disturbed by human activities (Yang et al., 2018; Liu X. et al., 2019), making its sea surface pCO₂ higher than that of the western part of the southern Yellow Sea.

The temporal variation of seawater pCO₂ in coastal regions resulted from the interaction of multiple processes, at least including temperature effect, air-sea CO₂ exchange, mixing processes and phytoplankton activity (Borges et al., 2006; Xue et al., 2016). Consistent with other studies of seawater pCO₂ (Xue et al., 2016; Hu et al., 2024; Li et al., 2023), we used Pearson correlation analysis to explore the factors affecting sea surface pCO₂ variations, where the square of the correlation coefficient r² represents relevance and the p-value of less than 0.01 in the analysis results represented 99% confidence level ( Figure 5 ). Consistent with the ~6 °C increase in SST, sea surface pCO₂ except for the low values from July 12 – 21 showed an increasing trend during the time series monitoring, and it was strongly positive correlated with SST (r² = 0.76, p<0.01) ( Figure 5A ), indicating the important influence of increasing temperature on the variations in pCO₂ during this period. In addition, the temperature driven pCO₂ values simulated according to the model proposed by Takahashi et al. (1993) were in good agreement with the observed values except for the low values during July 12 – 21 ( Figure 5A ), also suggesting the major control of SST on seawater pCO₂ during this period. However, the daily average temperature simulated pCO₂ value on July 25 was 34 μatm higher than that of the observed value, indicating influence of other factors.

By normalizing the pCO₂ values to the average SST, npCO₂ values were obtained to evaluate influence of other factors on sea surface pCO₂. SSS showed a decreased trend in this study, while npCO₂ did not evidently increase, and no correlation between npCO₂ and tide ( Figure 5B ), suggesting that the influence of terrestrial inputs with abundant pCO₂ was offset by other processes, such as biological activity and air-sea CO₂ exchange. During the whole observation period, the decreased trend of DO% and a ~30 µatm decrease in npCO₂ implied that biological activity was not an important factor controlling the seawater pCO₂ change. Considering the slight decrease in the level of npCO₂ during the overall period, the release of CO₂ from seawater to atmosphere in the air–sea exchange process may also influence the pCO₂ variations.

Since July 12, both the surface pCO₂ and npCO₂ showed an obvious decrease and the daily average of seawater pCO₂ and npCO₂ reached the lowest on July 16 – 17, with values of 563 μatm and ~488 μatm, respectively, which were 116 μatm of pCO₂ and 147 μatm of npCO₂ lower than those in July 12, and then the daily average pCO₂ and npCO₂ value increased to 669 μatm and 615 μatm on July 21, respectively. On the contrary, the daily average DO% reached ~102% on July 16 – 17 and then decreased. The npCO₂ exhibited a negative correlation with DO% during July 12 – 21 (r² = 0.51, p<0.01) ( Figure 5C ), indicating the important influence of biological activity on the variation in pCO₂ during this special period.

Based on satellite remote images, U. prolifera was observed to increase obviously on July 18 ( Figure 3E ), which was roughly consistent with the date of the seawater pCO₂ reduction, indicating the influence of U. prolifera bloom on carbonate system. The rapidly decreasing pCO₂ and increasing DO% in seawater was caused by the biological production of U. prolifera. It was reported that at the early bloom stage, photosynthesis of macroalga can quickly absorb dissolved inorganic carbon in seawater, leading to a decrease of pCO₂ in seawater (Li et al., 2014; Zhang et al., 2019). The cruise investigation at offshore area of Qingdao (Hu et al., 2024) also showed the decreased dissolved inorganic carbon and increased pH in seawater during bloom phase. Subsequently, as shown in satellite image on July 21( Figure 3F ), the amount of U. prolifera decreased, which may be caused by manual cleaning and sink of the decaying macroalga. Thus, the rising pCO₂ and decreased DO% in seawater after July 17 was related to the decomposition and respiration of decaying algae by microorganisms. Previous studies on the U. prolifera affecting the carbonate system in the Qingdao coastal area usually focused on the late bloom or after bloom period. These reports also found the release of CO₂ and reduced DO during the late bloom period (Deng et al., 2018; Xiong et al., 2023). However, previous studies were based on limited sample collection, and the effect of U. prolifera on carbonate system mostly examined vis incubation experiment, lacking in-situ continuous observations (Deng et al., 2018; Xiong et al., 2023). The time series observation in this work captured the significant effect of the continuous changing U. prolifera on seawater pCO₂, especially the influence on obvious decrease of seawater pCO₂ at early bloom stage, which confirms that the high-resolution measurements are essential for the study of carbonate system during algal blooms and provides a reference for the future study of the U. prolifera blooms influence on carbonate system.