CUI Shanggong, YU Xinsheng，ZHAO Guangtao

1. National Marine Data and Information Service, Tianjin 300171, China；

2.Key Lab.of Submarine Geosciences and Prospecting Techniques,MOE,College of Marine Geosciences,Ocean University of China, Qingdao 266100, Shandong province, China

Abstract: The oxygen fluxe across the sediment-water interface (SWI) in coastal region is a key measure to fully understand the regulation of biogeochemical cycles in an aquatic environment. However, studies on fluxes of dissolved oxygen in gravel beach are limited,because of the difficulty in sample collection and instrumentations deployment. In this study, benthic oxygen fluxes across rocky substratum in an intertidal zone of Huiquan Bay was estimated by using noninvasive eddy correlation techniques. A total of 10 burst measurements were analyzed.The oxygen flux fluctuated from-5.788 8±2.6 to+49.334 4 ±2.6 mmol O2 m-2/d were observed. The cospectra analysis showed that the oxygen flux at the frequency band between 0.093 and 0.279 Hz (at a period from 3.58 to 10.75 s)contributed 50.19% to the total spectrum on average. The results showed that the major contribution band moved to the high frequency region gradually and reached a steady state with increasing tidal flood. It is demonstrated that wave movement and wave breaking interaction resulted in the change of oxygen flux between gravel beach and shallow waters at the start and the end of a rising tide period, respectively. The eddy correlation techniques offer an efficient means for flux measurement over a gravel or mixed sand and gravel beaches.

Keywords: sediment-water interface(SWI), oxygen fluxes, eddy correlation,gravel beaches, in situ measurement

1 Introduction

Seabed is the tight coupling place of geochemical and biological processes between two phases:water column and sediment(Hopkinson and Wetzel, 1982). Not only sedimentwater interface(SWI) is recognized as important site for the mineralization of organic material and regeneration of nutrients in coastal environments(Middelburg et al., 2008),but it is an important region inhabited by high phototrophic biomass that contributes significantly to the sustenance of coastal food webs (Jahnke et al., 2000;Glud et al., 2009).Therefore, the oxygen fluxes across the boundary in sediment are the most commonly used proxy for benthic carbon mineralization and primary production (Canfield et al., 1993;Glud, 2008). Moreover, accurate quantification of the fluxes process are of great importance to understand the regulation of biogeochemical cycles and ecological system activities at various spatial and temporal scales (Berg et al., 2009).

Conventionally, oxygen-exchange fluxes across boundary in sediment are determined in situ by deployment of benthic chamber and oxygen microelectrode profiler (J ørgensen and Boudreau, 2001; Reimers et al., 2001; Wenzhöfer and Glud, 2004; Beer et al., 2005).In benthic chamber measurement, the flux is determined by measuring the change of concentration of dissolved oxygen with time in a fixed volume which is covered by benthic chambers. However, the enclosed chamber obstructs the natural hydrodynamic flow such as waves and currents and also could not represent the natural fauna density and behavior well(Berg et al., 2009). By measuring sediment O2profile with oxygen electrodes in situ, the diffusive fluxes of O2flux into the sediment can be calculated based on the O2concentration gradient at the sediment-water interface with Fick’s first law of diffusion(Glud, 2008). The disadvantages of O2profile approach are that the electrodes are fragile and the sophisticated equipment is required to control its movement precision. In addition,this method only provides the limited profile information of vertical points. Recently, eddy correlation(EC) method has been adopted into the aquatic environment for estimating fluxes across the sediment-water interface by simultaneously measuring high-frequency vertical velocity and the concentration of oxygen (Berg et al., 2003). It is a kind of non-invasive approach without disturbances of the sediment environment, ambient light conditions and the boundary layer flow during measurement. Compared with other methods, eddy correlation technique can estimate a much larger area on the seabed form 10 to 50 m2(Berg et al., 2007) with resolved fluxes down to at least ～1 mmol O2m-2/d(Berg et al., 2009).

In recent years, many efforts have been made to explore benthic fluxes with eddy correlation techniques, such as muddy deep-sea sediments(Berg et al., 2009), permeable sediments(Berg and Huettel, 2008), shallow freshwater(Brand et al., 2008), temperate seagrass beds (Hume et al., 2011) and Arctic sea-ice production(Long et al., 2012) etc.However, researches of oxygen fluxes on benthic hard surface have scarcely been reported. Gravel beaches, including mixed sand and gravel beaches, are typically found in most coast line around Qingdao city. There are relatively little research to study the biogeochemical processes in these area, mainly due to the difficulties associated with sample collections and field instrumentations deployments. Gravel beach are the main benthic habitats for micro or macro epifaunal organisms and algae, where oxygen changes are one of the main factor for their metabolic activities.

The main purpose of the present investigation is to evaluate benthic oxygen fluxes across the gravel beach by using the eddy correlation technique in the Huiquan Bay of Qingdao. Oxygen fluxes variations affected by waves and breakers during a rising tide period were exanimated and high-frequency bed fluxes change in the intertidal zone under wave condition are reported.

2 Methods

2.1 Field site

Huiquan Bay is a shallow coastal bay located on the east of Jiaozhou Bay, Qingdao city(Fig. 1). The bay has a trumpet-shaped mouth open toward the southwest. The maximum area is 1.24 km2at high water level. And the average water depth is 7 m. As the northern mountains blocked the WNW-N-NE offshore wind, The Huiquan Bay is mainly affected by the SSE Wind(Chang et al., 1992). It is a typical sandy and sandygravel mixed coast(Wang et al., 2007). Its semidiurnal tidal range is ～2.79 m. The current direction is clockwise at high tide and anti-clockwise at tide ebbed (Li et al., 1992;Zheng et al., 1998).

Usually, Chlorophyll a content in the top 2 cm of sediment is 1.73 - 5.71 μg/g(Wu,2005). The mean biomass of benthic microalgae in the intertidal zone in Qingdao is 0.55 μg/g(Jiang et al., 2006). The Huiquan bay also has abundant shellfish resources composed primarily of Littorina (L) brevicula (Philippi), Batillaria cumingi, Venerupis Philippinaram (Bai, 2011).

A sandy sediment field overlain by gravel and stones was chosen for eddy correlation system deployment on the east side of the bay(36°03'06.38"N, 120°20'23.36"E). It is the mid-tidal region of the intertidal zone and water depth is around 4 m for each tide.

2.2 Eddy correlation system

An acoustic doppler velocimeter (ADV) (Nortek AS, Norway) and a Clark-type oxygen microelectrode(OX-100,Unisense)were integrated onto a tripod frame. The ADV was used to measure and store the three velocity components (x-, y-, and z-directions) and the hydrostatic pressure at a maximum frequency of 64 Hz. The O2microelectrode was a Clark-type probe with outer tip diameter of ～90 μm, and 90% response times ＜2 s. In addition, a small CTD (OCEAN SEVEN 304Plus, IDRONAUT) was used to sample temperature, conductivity and pressure at 8 Hz.

Fig. 1 Map indicating study site, marked by blue star (a) and black star (b), was over an intertidal zone of Huiquan Bay in Qingdao

The ADV installed vertically facing the sediment surface and the microelectrode-tip positioned adjacent to the measuring volume of ADV(Fig. 2). The three velocity component and O2concentration data was typically recorded 24 cm above the sediment surface. The CTD was mounted approximately 60 cm above the sediment surface. Before deployment,ADV was programmed to measure in 20 min bursts, each consisting of a 15 min continuous data recording at 16 Hz followed by a 5 min pause. The electrode was polarized for at least 15 h prior to use and calibrated using a two-point calibration curve. It was settled continuous recording at 8 Hz for the CTD.

The system was deployed during low tide at 11:00 AM on June 4th, 2016. The frame was placed in the direction perpendicular to the dominant direction of tidal currents in order to avoid artifacts in flux estimates due to disturbance of current fields by the frame pods. Unfortunately, the datalogger broke down during 3.3 h later(10 bursts). Only data from 11:40 AM to 02 : 55 PM were available for analysis. These data covered a rising tide period basically according to the standard of North China Sea Marine Forecasting Center of the State Oceanic Administration.

2.3 Oxygen flux calculation

The mathematical expression for the vertical O2flux due to advection and molecular diffusion at any position in the water column at any point in time is defined as:

Fig. 2 Eddy correlation instrument and representative benthic photos. The instrument consisted of(A)an Acoustic Doppler Velocimeter (ADV), (B) a Clark-type oxygen microelectrode and (C) a CTD.The study site was a sandy sediment overlain by big stones covered with shellfish

where uzis the vertical velocity, C is the O2concentration, D is the molecular diffusivity of O2in water, and z is the vertical coordinate (Berner, 1980; Boudreau, 1997).

Due to turbulent advection is the dominant solute transport in the water column above the diffusive boundary layer, the diffusive term can be neglected. Then the Reynolds’decomposition is applied to both uzand C:

where the bars symbolize the averaging (Berg et al., 2003).

Fluxes were extr acted from the raw eddy correlation data in a multistep process.First, the spiking check was applied to dissolve oxygen (DO) and current velocity signal. It was critical not to lose real fluctuation that removing spikes caused by electronic noise and by sticking particles at the DO sensor. Second, we made a simple coordinate rotation for velocity signal of each burst with the ExploreV processing software(Nortek AS, Norway)so thatequals zero. Then all data recorded at 16 Hz were transformed to 8 Hz data in order to reduce noise while providing sufficient resolution to describe the entire frequency spectrum carrying the flux signal.

Because of the current directi on changes frequently in study site, the running averaging method became necessary to calculate the fluctuating terms uz′ and C′. The two means were calculated as running means over the minimum number of data points determined, such that an increase in number of data points did not result in any noticeable change in flux(Berg et al., 2003). The running means used in flux estimates were defined by data points between 2100 to 4200, equivalent to a time averaging interval between 4.375 to 8.75 min. After calculating turbulent fluctuations, time-shift and distance correction were proceeded by calculating cross correlate of the uz′ and C′ fluctuations,shifting the two data records relative to each other for each burst.

Due to our oxygen microelectrode response time was finite(＜2 s),we made frequency response enhancement to compensate for the flux lost by the equation:

where CouzC，measand CouzC，corrwere the measured and corrected cospectra of uz′c′, f was the frequency, and τCwas the 1/e response time of the DO electrode(Lorrai et al., 2010).Finally,oxygen fluxes,one of each 20 min measuring interval,were extracted from corrected 8Hz data.

3 Results

Examples of calculated fluxes, one for each 20 min segment of data, were shown in Fig. 3. Due to the data-logger problems during the measurements, the first 1.5 and last 4 bursts of data of depth and temperature were missing.

During measurement, the water depth reached 2.80 m at high water level. The water temperature had sustained to decline from 17.58 °C to 16.82 °C in an hour and a half.The mean velocities of vertical currents ranged from 0.0 to 0.7 cm/s.

Fig. 3 The full eddy correlation data: (A) Vertical velocity component. (B) O2 concentrations recorded with a microelectrode. (C) Water depth recorded with CTD. In last four bursts ADV pressure data was used to replace and the yellow line represented fitting values. (D) Temperature recorded with CTD. (E)Derived fluxes. Positive fluxes indicate a release of oxygen from the sediment (error bars represent SE).

It was observed that O2concentration variations were ranged within the record from 278.20 μmol/L to 252.89 μmol/L. During the initial period, the O2concentration decreased rapidly and was recovered alter around 12:10. The reason for this could be the gaseous oxygen in the sand was dissolved in the water when the tide rises. After that, the O2concentration decreased continually with increasing water depth (Fig. 4).

Fig. 4 O2 concentrations variation with increasing water depth during tide rising

Tab. 1 Data of overlying water parameters including O2 concentration, O2 flux, temperature, water depth and current velocity during each burst. Mean values are indicated in parentheses. Positive flux represents sediment release

Tab. 1 summarized specific mean values of calculated O2fluxes and associated current velocities. The oxygen exchange fluxes observed in this study were comparable to those observed in reported studies in sandy sediment at the same Bay by using the benthic chamber method(range: -61.146 ±10.4 to + 14.998 ±1.2 mmol O2/m2/d) (Zhang et al., 2005). However, there was obvious regularity in fluxes. In burst 1, the flux was negative due to the low water level at the beginning of a rising tide. The DO concentration in the surface of water volume was considerably higher than those in the surface of sediment. In burst 2 and 3, the flux became positive. It was attributed to dissolution of gaseous oxygen. When the sediment was covered by water, gaseous oxygen in pore sediment was dissolved in water so that the DO concentration at the bottom was higher.After that, the flux decreased gradually to a negative value and then recovered to +49.3 ±3.7 mmol/m2/d. This could be caused by several mechanisms. First, with the rising tide,the distance between measurement point and the water-air interface expanded. The DO concentration in the surface was higher than those on the underside so causing a downward vertical flux through a measurement point. The second was the influx of oxygen-enriched water caused by wave. The wave could cause the erosion of DO concentration gradients and increase the DO flux in the downward direction(Holtappels et al., 2013). Despite low carbon content, permeable sands could have high O2turnover rates due to an efficient entrapment and recycling of particulate carbon(Huettel et al., 1996,2007; Cook et al., 2007). The capability for such sands to work as biocatalytic filters facilitating an efficient carbon turnover does, and was maintained by an occasional wave-induced high-energy flow (Huettel and Webster, 2001).

As calculated cumulative fluxes whose slope of trend was equal to O2flux(Fig. 5),cumulative fluxes did not show a nearly linear trend within every burst. It could represent extremely unstable conditions in the study site(Berg et al., 2009). This may indicate such continuous flushing event driven by the concurrent increase in current velocity(Berg et al.,2013). So our derived mean fluxes were statistically representative of all eddy sizes that contribute to the flux and it was normal in the rising tide condition.

Fig. 5 Cumulative fluxes with volatility in the rising tide condition (burst 3-5)

4 Discussion

4.1 The low oxygen flux values

The measurements in this study indicated that the oxygen flux was extremely unstable during rising tide period. It changed from -5.7888 to + 49.334 4 mmol O2m-2/d(average: +11.387 6 mmol O2m-2/d). These values were lower than measured ones in restored eelgrass meadow (Hume et al., 2011) and shallow freshwaters (McGinnis et al.,2008). This could occur by several possible reasons. First, it was a gravel and sandy mixed field overlain by big stones where sediment layer was thin. The volume of sediment accumulation was small with low carbon content. Second, sediment re-suspension caused by wave action could block photosynthesis by reducing light availability (Lawson et al.,2007; Hume et al., 2011).

4.2 Sensitivity of spectral flux contributions

The normalized cumulative cospectra, spectra density of the oxygen concentration and the vertical velocity were evaluated respectively. Typical results of the first and sixth burst were shown in Fig. 6. For curve of burst 1, the frequency band between 0.093 to 0.279 Hz(at a period from 3.58 to 10.75 s) indicated a major contributor to the total spectrum. The temporal interval was coincident with mean wave period between 5.4 to 11.0 s(Chang et al., 1992). As pointed out by Lorrai C. et al(2010), the eddy timescale was 0.2～10 s in high turbulence regions (the horizontal velocity at 1 m above sediment U1m= 20 cm/s). Our results agreed quite well with the reported results. The contribution percentage of these components to the total power spectrum density of uz′ and C′ was 20.82% and 23.64%, respectively. However, the contribution of high-frequency components(＞1 Hz) to the total power spectrum density was low. Noise contaminations contained in the high-frequency band of the velocity and oxygen concentration was presumably random and non-correlated(white noise), resulting in the negligible contribution to the total flux(Kuwae et al., 2006). For curve of burst 6, the contribution of high-frequency components was significantly stronger in the flood tides, although the frequency band between 0.1 to 0.3 Hz was still outstanding. In addition, the energy of frequency between 0.4 to 0.62 Hz(at a period from 1.61 to 2.50 s) became a new ancillary contributor. In the cumulative cospectra display, the two curves had increased when frequencies decreased. This phenomenon revealed that the contribution of high-frequency components was gradually increasing with water level rising. These characteristics represented turbulent mixing generated by wave action, which showed a steep part in the spectra within a narrow frequency band. According to statistical information of wave period in Huiquan Bay(Chang et al., 1992), the wave dominated the hydrodynamics there.

4.3 Transformation of spectral flux contributions

To investigate the rate of change, the cumulative cospectra of every burst was compared(Fig. 7). It’s obvious that the first burst showed a concave curve and the frequency band between 0.093 to 0.279 Hz was a major contributor. For the last three bursts, the pattern became a coincident convex curve. It illustrated that effect of high-frequency components could enhance rapidly in the first two hours of the rising tide period and reach a steady later on. The percentage of frequency band which was greater than 0.8 Hz increased from 20.17% to 71.40%. The measured motions during high-tide period were likely caused by turbulence, leading to relatively large amplitude fluctuations in c and w.

Fig. 6 Typical spectra density results of the first and sixth burst

4.4 Error analyses of time-shift and distance correction

Fig. 7 The cumulative cospectra of burst 1,3,7 and 10. It illustrated that effect of high-frequency components could enhance rapidly in the first two hours during rising tide period and reach a steady later on

There was inevitably some delay in the DO signal with respect to velocity measurement due to the different frequency responses. The mismatch between the sensor tip and the ADV sample volume also caused some time offset, depending on flow velocity and direction. The cumulative cospectrum would contain a time-shift-dependent artificial flux contribution if the correction would not be applied. Through cross correlating the uz′ and C′ fluctuations, the resultant time shift could be quantified and the two data records could be shifted relative to each other.

The time-shift correction may not be constant throughout the time series because of the varying currents. Therefore, this correction should be performed on segments with approximately steady conditions. In our experience, the flux loss due to the time-shift correction of burst 1 and burst 6 was 17.11% and 0.21%, respectively (Fig. 8).

It should be noted that the oxygen fluxes is based on adequate time integration. This is especially important for permeable sediments because of their dynamic fluxes(Glud,2008). A momentary stimulation of oxygen uptake cannot be used as a direct measurement of an equivalent increase in carbon mineralization. Because release of anoxic pore water and oxidation of stored reduced compounds will likely contribute significantly to the measured oxygen flux(Berg et al., 2013). So it is needed to obtain long time series in order to resolve the correct average O2exchange for heterogeneous environments.

Fig. 8 Time-shift corrected cumulative cospectrum (solid line) compared to the cumulative cospectrum of the original(nonshifted) data showing an artifact contribution at higher frequencies (dotted line).(Left) in the first burst. (Right) in the sixth burst

Because of instrumentation problem, the data set recorded for this field measurement is limited. A long period deployment with multiple in situ sensors will be considered in future studies.

5 Conclusions

In summary, in situ measurement of oxygen fluxes at gravel beach has been made in Huiquan Bay. A total of 10 burst measurements were analyzed. The oxygen flux fluctuated from -5.788 8 ±2.6 to +49.334 4 ±2.6 mmol O2m-2/d was observed.The results indicated a strong variation in benthic DO flux in intertidal zone driven primarily by wave. The comparison of DO flux variations in different rising tide period was conducted and revealed that the tidal wave has significantly impact on benthic DO flux changes. Spectra analysis indicated that the frequency band between 0.093 to 0.279 Hz (at a period from 3.58 to 10.75 s) was a major contributor to the total spectrum. The results suggest that the wave and wave breaking interaction dominated the hydrodynamics at the start and end of a rising tide period.

Acknowledgemets

We thank Jiang Zike, Zhang Wei and Wang Yanyan for helping in field works. We grateful for the technical assistance of Dr. Wang for constructing the observation frame.Funding for this work was supported by the National Natural Science Foundation of China(Nos. 41276089, 41176078) and the National High-technology research and development Program of China (‘863’Program) (Nos. 2012AA09A20103, 2009AA09Z201).