1. Introduction

Red Noctiluca scintillans (RNS) is the primary species causing red tide in coastal waters in China [1–3], and its distribution is common in coastal waters throughout the world [4–7]. RNS is found over a wide temperature range of $\mathrm {10-25^{o}C}$ and high-salinity coastal regions of China seas at different seasons, for instance, the Liaodong Bay in summer [8], the Jiaozhou Bay and adjacent coastal Yellow Sea throughout the year [9,10], the East China Sea in spring, summer and fall [3,11], as well as the Daya Bay and Dapeng Bay in spring [1,12]. Red tides such as those caused by RNS can lead to damage to ecosystems and substantial economic losses [3,13,14]. It is essential to detect RNS bloom using a timely manner as well as to monitor the abundance of RNS for the early warning of red tide with a suitable proxy.

Collecting water samples to identify phytoplankton species by microscopy is a traditional and reliable method to obtain information on their abundance . However, it cannot be applied to large-scale field monitoring of phytoplankton bloom events because of the limitation of the high associated cost. Alternatives that can be deployed widely on different kinds of platforms are needed in field monitoring . For field detection/monitoring of RNS blooms, many efforts, such as shipboard inherent optical property measurements, biogeochemical real-time monitoring buoys, and airborne or satellite ocean color remote sensing, have been made, and some results have been achieved [3,15–19]. These methods make full use of some proxies, such as the bio-optical properties, reflectance spectrum or ecological factors when monitoring RNS blooms. However, these methods or proxies have their own limitations. It is reported that absorption spectral shape of RNS varies between stations, which is probably related to their gut content and to the kind of food they are eating since RNS lack their own pigmentation [15], which makes it difficult to distinguish RNS by absorption spectrum. For medium and small-scale RNS in the field, there is no significant difference between the spectra of RNS and other phytoplankton , therefore, it is difficult to distinguish them using spectral methods including inherent optical properties and reflectance . Moreover, for inherent optical properties, RNS abundance is difficult to quantify by using absorption and scattering properties before visible red tide outbreaks [15,16]. Similarly, ocean color remote sensing is only suitable when the RNS bloom patches are large and the RNS concentration is sufficiently high near the surface and is limited by the satellite overpass time and spatial coverage due to clouds [19]. Obviously, remote sensing is difficult to use as a normal tool for estimating in situ RNS abundance. Therefore, the method as well as a relevant proxy for monitoring in situ RNS abundance needs further development and research.

Recently, bioluminescence has been utilized as an optical proxy to monitor marine organisms in the field [17,18,20]. Different types of marine bioluminescence are generated chemically by a wide variety of marine organisms, from bacteria to large squids and fishes [21], and can be observed visually during a voyage or at port. Currently, quantitative bioluminescence characteristics provided by field instruments have been used to discriminate between taxa directly [20], including phytoplankton and zooplankton. However, for field monitoring of the abundance of dinoflagellate species, the quantitative relationship between bioluminescence and abundance needs to be further explored. Bioluminescence usually occurs at the sea surface, such as glowing wakes, and is related to phytoplankton species and abundance [22–25]; this type of bioluminescence is caused by mechanically disturbing blooming phytoplankton, notably the microscopic dinoflagellate Noctiluca scintillans. Bioluminescence has become a proxy of RNS abundance as a supplement for normal long-term field observations [17]. In addition, Li et al. [26] showed the geographic pattern of bioluminescence intensity and the environmental driving factors including temperature, salinity and chlorophyll, as well as their correlation in the Yellow Sea and Bohai Sea (YSBS), and the focus was on the effects of hydrological and chemical parameters on bioluminescence rather than bioluminescent organisms [26]. Therefore, the spatial distribution characteristics of bioluminescence and bioluminescent dinoflagellates are worth further discussion. The quantitative relationship between RNS abundance and bioluminescence in the YSBS also needs to be analyzed with bioluminescence flash kinetics.

In this research, comprehensive in situ data, including the abundance of multiple dinoflagellate species and bioluminescence profiles, were observed simultaneously from a summer cruise in the YSBS, and their spatial distributions were analyzed. Then, the relationship between bioluminescence and dinoflagellate species is explored in detail with these in situ data, combined with the bioluminescence flash kinetics of dinoflagellate individuals . Ultimately, a novel method for estimating RNS abundance by bioluminescence is proposed, which can be used in field monitoring.

2. Data and methods

2.1 Cruise

Under the support of the National Natural Science Foundation of China, in situ data including bioluminescence and dinoflagellate abundance were obtained simultaneously at 94 stations aboard the R/V "DONGFANGHONG 2 " during July and August 2018 in the YSBS. The distribution of sites is shown by black circles in Fig. 1. Bioluminescence data of the entire profile from surface to bottom were measured by the Underwater Bioluminescence Assessment Tool (UBAT, WET Labs Inc.) [27] with a falling speed of $\mathrm {0.2~m~s^{-1}}$ . To count the abundance of individual dinoflagellates, seawater samples were collected at the surface, subsurface, middle and bottom layers according to the water depth of the station. To obtain bioluminescence data of each layer, the bioluminescence profile data within sampling depth are averaged by $\mathrm {1~m}$ interval. In addition, RNS was observed in 76 of 252 samples. Bioluminescence measurements of individual dinoflagellates were also performed in the laboratory on board (for details, see section 2.3). Moreover, bioluminescent wakes were observed visually at several stations and their surroundings at night.

 

Fig. 1. Water depth of the study area and distribution of in situ sites in the YSBS. The red characters (A, B and C) are marked to demonstrate typical profiles of the MBP and individual dinoflagellate abundance (see section 4.2). The red line is Transect 1 (see section 3). The locations of the Yellow Sea and Bohai Sea are labeled by their name.

Download Full Size | PDF

2.2 Postprocessing of bioluminescence data measured by UBAT

Bioluminescence data were measured by UBAT [27]. UBAT can receive more than $95\%$ of photons from $\mathrm {430{~nm}}$ to $\mathrm {700{~nm}}$ and provide measurement of maximum bioluminescence potential (MBP, in ${\mathrm {photon~s^{-1}~L^{-1}}}$) stimulated mechanically by the internal pump [28]. Bioluminescence data are sampled under a $\mathrm {60~Hz}$ sampling rate to measure the short time scale evolution of the bioluminescence signal, which is important for analyzing the flash kinetics of bioluminescent dinoflagellates [20]. The UBAT data file provides the coefficient for detector gain (in $\mathrm {photon~s^{-1}}$), rotate speed of pump (in rpm), average bioluminescence intensity (in $\mathrm {photon~s^{-1}}$), and $\mathrm {60~Hz}$ digitized raw analog to digital (A/D) counts of bioluminescence intensity.

The $\mathrm {60~Hz}$ digitized raw A/D counts can be converted into bioluminescence intensity (in $\mathrm {photon~s^{-1}}$) using Eq. (1).

The average bioluminescence intensity corresponds to the averaged value of these 60 bioluminescence intensities over one second. To derive the MBP, the average bioluminescence intensity must be transformed to MBP by Eq. (2).

The Flow rate, in $\mathrm {L~s^{-1}}$, equals the rotate speed of pump (in rpm) in the UBAT data file multiplied by the flow rate calibration coefficient that is specific to each UBAT and is provided by WET Labs on the calibration sheet.

The MBP, in ${\mathrm {photon~s^{-1}~L^{-1}}}$, is defined as averaged bioluminescence intensity per unit time and unit volume. And the unit of bioluminescence intensity per dinoflagellate cell in a volume should be $\mathrm {photon~cell^{-1}~s^{-1}}$. Thus, the abundance of dinoflagellates, in $\mathrm {cell~L^{-1}}$ can be obtained by the ratio of the two parameters. So the abundance of dinoflagellates is estimated by MBP rather than bioluminescence intensity.

2.3 Onboard laboratory measurements of bioluminescence flash kinetics

To characterize the bioluminescence of dinoflagellate individuals , UBAT was used to detect flash kinetics on a single living cell in a laboratory tank onboard the R/V "DONGFANGHONG 2". For onboard laboratory bioluminescence measurements, in situ dinoflagellate samples were collected by CTD sampling bottles at stations the same as the UBAT measurement. A single living cell of dinoflagellate individuals was immediately identified and separated by microscopy, and then transferred into a 25 L tank filled with prefiltered ($\mathrm {20~\mu m}$) seawater. Prefiltered seawater was prepared prior to each laboratory measurement by collecting seawater from the sea surface at the same station and filtering. The tank was rewashed with the prefiltered seawater every time before performing measurements, and new, fresh and prefiltered seawater was used for each living cell. Filtered seawater gave a maximum bioluminescence intensity of $\mathrm {1-2\times 10^{7}~photon~s^{-1}}$, indicating bioluminescence background signal in the underwater UBAT. UBAT was prewashed with pure water and submerged into the tank to examine species by species, and the settings were the same for measuring all species and all profiles. Living cells were gently placed into the tank in front of the inlet of the UBAT. The time required for each living cell to sink into the seawater and the appearance of maximum bioluminescence intensity were recorded together as ancillary data to exclude occasional measurements because the bioluminescence efficiency of some dinoflagellates is relatively low (see section 5) and their bioluminescence data may be affected by instrument noise of UBAT. For a certain dinoflagellate species, its bioluminescence flash kinetics is obtained by averaging all valid measurements of each cell. Although the measurements used to average were not obtained at a station, due to their similar bioluminescence characteristics within certain regions, the averaged flash kinetics curve could represent the bioluminescence characteristics of dinoflagellate individuals [20]. The relationship across all the data shows that the data can be aggregated across regions. In addition, according to obtained flash kinetics, the corresponding species is identified as bioluminescent dinoflagellate species contributing to in situ bioluminescence.

2.4 Abundance of dinoflagellate individuals in the laboratory

During the measurement of MBP, 2 L seawater samples from different layers, including the surface, subsurface, middle and bottom layers, were collected simultaneously . Samples were concentrated to 250 mL using a $\mathrm {20~\mu m}$ sifter and immediately poured into 2% formaldehyde solution for fixation on site. After the samples were brought back to the laboratory, they were concentrated to 25 mL. Then, the 25 mL sample was allowed to settle for a minimum of 24 h. Species identification and counting were conducted under an inverted biological microscope. After the number of cells in the sample was obtained, the abundance of dinoflagellate individuals, ABUND in $\mathrm {cell~L^{-1}}$, was calculated by the number of cells divided to 2.

2.5 Statistical parameters

The empirical statistical relationship between MBP and RNS abundance is established using in situ data. The least square method was used to analyze the relevance between MBP and RNS abundance. The goodness of fit is described by the determination coefficient ($\mathrm {R^{2}}$) and root mean square error (RMSE), which can be calculated by Eqs. (3) and (4), respectively.

3. Model to retrieve RNS abundance based on the MBP

For prewarning of red tide, accurate high abundance data are more important than low abundance data because the risk of red tide is low for low dinoflagellate abundance. According to the reported RNS abundance leading to red tide [2], level-4 warning of red tide should be released and relevant areas should be monitored every 3 days when RNS abundance exceeds $\mathrm {500~cell~L^{-1}}$. In addition, the abundance of $\mathrm {100~to~500~cell~L^{-1}}$ is a prewarning threshold.

RNS is one of the dominant dinoflagellate species in red tide in the YSBS, while Pyrocystis, which has high bioluminescent ability, was not found in our investigation. In the YSBS, compared with other bioluminescent dinoflagellates, RNS has the highest bioluminescence efficiency (see section 5). In other words, the contribution of other dinoflagellates to the MBP is much less than that of RNS. Therefore, the method of estimating RNS abundance using MBP is feasible in the YSBS.

To derive the RNS abundance using the MBP, the data pairs with MBP and counted RNS abundance are selected to establish empirical statistical relationship under the criteria that MBP exceeds $10^{8}$ photon $\textrm{s}^{-1}$ $\textrm{L}^{-1}$, RNS abundance exceeds $0.5$ cell $\textrm{L}^{-1}$ and the proportion of RNS exceeds 10%, which could be considered to represent valid bioluminescence measurements dominated by RNS. In fact, $\mathrm {0.5~cell~L^{-1}}$ represents one cell in the $\mathrm {2~L}$ bottle (see section 2.4). After data matchup and quality control, there were 35 of 76 RNS abundance values that could be used to establish the relationship. It needs to be explained that the excluded data can be regarded as the two situations, there is RNS abundance but no MBP measurement, and there are several of RNS abundance values accompanied by high abundance of other dinoflagellates. The empirical relationship between RNS abundance and MBP is represented by Eq. (5) on a logarithmic scale.

For illustrating the relationship between MBP and RNS abundance, Fig. 2(a) shows the log-log plot for avoiding a correlation that is biased to the denser populations, and Fig. 2(b) shows the linear-linear plot for physical reality. As shown in Fig. 2(a), a high $\mathrm {R^{2}}$ value of approximately 0.82 and low RMSE (0.39) indicate that RNS abundance is highly related to the MBP. There is a good linear relationship between the MBP and RNS abundance on a logarithmic scale. In addition, two red circles with RNS abundance above $\mathrm {0.5~cell~L^{-1}}$ and MBP below $\mathrm {10^{9}~photon~s^{-1}~L^{-1}}$, which are obtained in the bottom layer and not included in the relationship, are also shown in Fig. 2(a), indicating that there are times when RNS are present, but not luminescing. Although there are some outliers, the relationship between RNS abundance and the MBP converges from small values to large values. Thus, the regression model can be used to estimate RNS abundance from the MBP.

RNS abundance can be graded in theory according to the order of magnitude with the MBP. In detail, RNS abundance is less than 1 $\mathrm {cell~L^{-1}}$ when MBP is less than $\mathrm {10^{10}~photos~s^{-1}~L^{-1}}$. RNS abundance from 1 to 10 $\mathrm {cell~L^{-1}}$ corresponds to MBP from $\mathrm {10^{10}}$ to $\mathrm {10^{11}~photos~s^{-1}~L^{-1}}$. When RNS abundance is between 10 and 100 $\mathrm {cell~L^{-1}}$, MBP is between $\mathrm {10^{11}}$ and $\mathrm {10^{12}~photos~s^{-1}~L^{-1}}$. RNS abundance is larger than 100 $\mathrm {cell~L^{-1}}$ when MBP is larger than $\mathrm {10^{12}~photos~s^{-1}~L^{-1}}$. This classification also coincides with the bioluminescent efficiency of phytoplankton species (see section 5). Therefore, it can be inferred that bioluminescence can be used as an indicator of the level of RNS abundance.

Figure 2(c) shows the comparison of estimated and counted RNS abundance. Additionally, the measurement data from Transect 1 (referred to sites at $\mathrm {33^{o}N}$, as shown in Fig. 1) were measured in July 2018 to verify the accuracy of the regression relationship. The transect is composed of four stations. Figure 3 illustrates the MBP of this transect from the surface layer to the bottom layer. MBP decreases with increasing depth. In the transect, MBP ranges from $\mathrm {8.6\times 10^{9}}$ to $\mathrm {1.8\times 10^{12}~photon~s^{-1}~L^{-1}}$, and RNS abundance is between 5 and 320 $\mathrm {cell~L^{-1}}$. High RNS abundance occurs at the location and depth where the MBP is high. RNS abundance estimates at the sampling depth were calculated by the MBP according to Eq. (5). The estimated RNS abundance is approximately 2 to 305 $\mathrm {cell~L^{-1}}$ and has a similar distribution with in situ RNS abundance. Comparing in situ and estimated RNS abundance, the average relative error is approximately 41%, and more than half of the estimates are within 40% of the relative error, where a large error corresponds to low RNS abundance. When the in situ RNS abundance exceeds $\mathrm {10~cell~L^{-1}}$, the average relative error in this transect decreases to approximately $25\%$.

 

Fig. 2. The relationship between MBP and RNS abundance on the $\mathrm {log_{10}-log_{10}}$ scale (a) as well as linear-linear scale (b). The black circles are used to fit rather than red circles. Comparison of estimated and counted RNS abundance (c). N is the statistical number. It is noted that $\mathrm {0.5~cell~L^{-1}}$ is one cell in the $\mathrm {2~L}$ bottle.

Download Full Size | PDF

 

Fig. 3. Comparison of the MBP, in situ and estimated RNS abundance of Transect 1. The MBP of four stations is represented by the contour map. In situ and estimated RNS abundance are shown by red dots and black circles, respectively. The symbol size indicates RNS abundance, as shown in the legend. The longitude of each column of symbols corresponds to the station position.

Download Full Size | PDF

For the proposed method, high RNS abundance would hardly be underestimated and its error is relatively low. Therefore, the magnitude of the estimated RNS abundance and its spatial distribution is acceptable and practical, which is possibly used to provide an effective early warning of RNS red tide.

4. Optical proxy from bioluminescence manifested by spatial distribution

4.1 Spatial distribution in the surface waters of the YSBS

The distributions of bioluminescent dinoflagellate species and MBP in the surface waters of the YSBS need to be considered first. The flash kinetics of five dinoflagellate species, including RNS, Dinophysis, Protoperidinium, Ceratium furca, and Ceratium horridum were obtained by onboard laboratory measurements (see section 5), so the distributions of the five dinoflagellate species are shown in this section. The total counted dinoflagellate abundance and the proportion of the five dinoflagellate species are shown in Fig. 4(a) and 4(b), respectively. The MBP is also shown in Fig. 4(b) to illustrate the relationship between MBP and dinoflagellate species. The relationship could be qualitatively analyzed by Fig. 4.

 

Fig. 4. Distribution of counted dinoflagellate species abundance (a) and proportion of five dinoflagellate species, and MBP (b) in the surface waters of the YSBS. (a) Total abundance of all identified dinoflagellate species. (b) The shape and size of the symbols represent the dinoflagellate species and the proportion of dinoflagellate individuals, respectively. The contour map shows the MBP.

Download Full Size | PDF

The relationship between MBP and dinoflagellate species is complex. First, generally speaking, the high MBP area is consistent with the high abundance area, which indicates that it is possible that the more bioluminescent dinoflagellates is, the larger MBP is. In addition, it is noted that certain regions do not have the consistency, such as the Bohai Sea, which has a high abundance and relatively low MBP, and the center of the Yellow Sea, which has a high MBP and relatively low abundance. It demonstrates that signal contributed from each specie varies very significantly. In other words, the proportion of dinoflagellate species plays a key role in the level of MBP. Additionally, the bioluminescent efficiency of dinoflagellate individuals varies greatly. Finally, by comparing averaged MBP in the Bohai Sea ($\sim \mathrm {9.45\times 10^{9}~photon~s^{-1}~L^{-1}}$) with a high proportion of Dinophysis and the southern Yellow Sea ($\sim \mathrm {1.61\times 10^{11}~photon~s^{-1}~L^{-1}}$) with a high proportion of RNS, it is obvious that MBP is relatively high when RNS is the dominant species. Combining the total abundance with the proportion of dinoflagellate individuals , it is likely that MBP is positively correlated with RNS abundance, even though other dinoflagellate species exist at the same time, which is consistent with Kim et al. [17]. Therefore, a quantitative relationship can be attempted to infer from RNS first (detailed discussion in section 3 and 5).

4.2 Vertical distribution

Although observed bioluminescence due to dinoflagellate blooms usually occurs at the sea surface, according to MBP profiles, there is probably bioluminescence at different depths. Figure 5 shows typical profiles of MBP and dinoflagellate abundance at three sites in the southern Yellow Sea (sites see Fig. 1). Among these sites, the magnitude of MBP near the $\mathrm {20~m}$ layer is close to that at the surface, which indicates that the bioluminescence phenomenon may occur below surface. Moreover, both MBP and dinoflagellate abundance exhibit covariation trends in the vertical direction, so MBP could reveal the dinoflagellate abundance. Comparing RNS abundance with $\mathrm {MBP/10^{10}}$, the order of their magnitude is likely the same. From Fig. 5, the abundance of RNS and other dinoflagellates, such as Ceratium furca and Protoperidinium, may affect the magnitude of MBP. However, the comparison between Figs. 5(a), 5(b) and 5(c) reveals that only high RNS abundance corresponded to high MBP, while low RNS abundance corresponded to low MBP, such as $\mathrm {10^{2}~cell~L^{-1}}$ corresponding to $\mathrm {10^{12}~photon~s^{-1}~L^{-1}}$ and $\mathrm {1~cell~L^{-1}}$ corresponding to $\mathrm {10^{10}~photon~s^{-1}~L^{-1}}$. Therefore, RNS is the most likely species leading to high bioluminescence in this area.

 

Fig. 5. Vertical distribution of MBP and individual dinoflagellate abundance in the southern Yellow Sea. Figures 5(a), 5(b), and 5(c) correspond to A, B, and C marked in Fig. 1, respectively.

Download Full Size | PDF

Taking a few seawater samples at each site into account, it is difficult to illustrate the complete vertical distribution of dinoflagellate abundance. To obtain the vertical distribution of bioluminescent dinoflagellates, the characteristics of MBP profiles are worth exploring. As shown in Fig. 6, the MBP profiles measured at each site are plotted in the same axes, and the color of line represents the depth of the site. Combined with the water depth of the YSBS shown in Fig. 1 and the surface MBP shown in Fig. 4(b), some characteristics of vertical MBP profiles could be found as follows.

 

Fig. 6. Distribution of MBP profiles measured at each site in the YSBS. The line color represents the water depth at the site.

Download Full Size | PDF

In general, the magnitude of the MBP could change by several orders in the vertical direction. For instance, at a certain station with a water depth greater than $\mathrm {60~m}$, the magnitude of the MBP decreases from $\mathrm {10^{12}~photon~s^{-1}~L^{-1}}$ at the surface to $\mathrm {10^{7}~photon~s^{-1}~L^{-1}}$ at the bottom. However, there are other circumstances where the MBP maximum occurs at the middle or lower layer rather than the upper layer, regardless of water depth. In addition, the lines in Fig. 6 are all zigzags indicating that there is no completely uniform MBP throughout the water column , which is related to some relatively large and sparse particles in vertical direction.

The magnitude of the MBP at the surface varies in different regions, and the range of the variation in the vertical MBP varies in different regions. For the sites with depths less than $\mathrm {20~m}$, in the Bohai Sea, the magnitude of vertical MBP ranges from $\mathrm {10^{8}}$ to $\mathrm {10^{10}~photon~s^{-1}~L^{-1}}$, and in other areas, the MBP could reach $\mathrm {10^{11}-10^{12}~photon~s^{-1}~L^{-1}}$. For the sites with depth ranges of $\mathrm {20-50~m}$, the variation in the vertical MBP does not exhibit a regular pattern, but the most obvious subsurface maximum of the MBP appears at these sites. Moreover, the MBP in the lower layer could be reduced to the order of $\mathrm {10^{7}~photon~s^{-1}~L^{-1}}$ when the water depth exceeds $\mathrm {20~m}$, especially in the Yellow Sea with water depth greater than $\mathrm {50~m}$. The variation in the MBP in the lower layer could change by at most four orders for the sites in the Yellow Sea where the depth exceeds $\mathrm {50~m}$.

Both the magnitude and variation in the vertical MBP are very large. Taking the potential vertical migration of dinoflagellates [29] into account, bioluminescence at the sea surface may also be related to the vertical distribution of the MBP.

5. Flash kinetics of bioluminescent dinoflagellates in the YSBS

In the YSBS, eight bioluminescent dinoflagellates, such as RNS [17,26], Dinophysis, Gonyaulax ploygramma, Ceratium furca, Ceratium horridum, Ceratium fusus, Protoperidinium and Alexandrium, were found in seawater samples. Among them, the bioluminescence flash kinetics of RNS, Protoperidinium, Dinophysis, Ceratium horridum, and Ceratium furca were obtained from onboard laboratory measurements, while the other three dinoflagellates simply did not luminesce. Based on the above comparison between the distribution of the in situ MBP and counted dinoflagellate abundance, there is a strong correlation between RNS abundance and the MBP, while the correlations between the abundance of other dinoflagellate and the MBP are very poor [17] . In this section, the reason why bioluminescence is more feasible to estimate the abundance of RNS rather than other bioluminescent dinoflagellates is attempted to be explained by analyzing the bioluminescence flash kinetics of these dinoflagellates.

On the one hand, the existence of RNS is a necessary condition for estimating the RNS abundance with the MBP. In past investigations [30,31], RNS, as one of the primary bioluminescent dinoflagellates in the YSBS, especially in the Yellow Sea, was found to have higher abundance and bioluminescence efficiency than other dinoflagellates. In our investigation, RNS ranked the third in terms of species dominance among the eight dinoflagellates in the YSBS. These dinoflagellates are listed in descending order of abundance as follows: Ceratium fusus, Ceratium horridum, RNS, Ceratium furca, Gonyaulax ploygramma, Dinophysis, Protoperidinium, and Alexandrium. Thus , using the MBP to estimate RNS abundance is practical in the YSBS.

On the other hand, the bioluminescence efficiency of dinoflagellates is important, which could be represented by the potential maximum bioluminescence intensity of a single cell, and the bioluminescence efficiencies of the eight dinoflagellates reported by different researchers are listed in Table 1. However, due to the lack of a unified observation protocol for bioluminescence, the difference of bioluminescence efficiency affected by factors including research purposes, experimental conditions, sensors and bioluminescence units, are difficult to be quantitated. And the results are also difficult to be converted to each other under different conditions . In addition, taking the impacts of local eco-environmental factors specific to each area into account [7,32], dinoflagellates may have different bioluminescence abilities in different seas [24,33,34]. Therefore, it is suitable that using these previously reported results as qualitative reference to indirectly confirm the bioluminescence flash kinetics analysis described below.

Table 1. Bioluminescence efficiency in the literature.

View Table | View all tables in this article

To illustrate the bioluminescence characteristics of dinoflagellate individuals in this study, bioluminescence flash kinetics curves were obtained and analyzed. Figure 7 shows the bioluminescence flash kinetics curves for five dinoflagellate species : RNS, Protoperidinium, Dinophysis, Ceratium horridum, and Ceratium furca. The values of their characteristics are tabulated in Table 2. As shown in Fig. 7, for these dinoflagellates, bioluminescence intensity rises rapidly until the time to reach its peak and decays slowly to the end of bioluminescence. Although the difference of their maximum bioluminescence intensity is large, it is noted that the shape of their kinetics curves is similar and the range of their bioluminescence time is overlapped. Taking $\mathrm {60~Hz}$ sampling rate of the UBAT in account, the time to reach peak of these dinoflagellates, which are $\mathrm {0.033~s}$, $\mathrm {0.05~s}$, $\mathrm {0.067~s}$ and $\mathrm {0.083~s}$, are too close to discriminate in field measurements. The lifetime exponent represented by the total time of bioluminescence seems to be positively correlated with maximum bioluminescence intensity. However, in terms of magnitude, its maximum variation $\mathrm {0.4~s}$ is far less than that of the maximum bioluminescence intensity. There is no significant difference in other kinetic features including time to reach peak or lifetime exponents for the five dinoflagellate species. Thus, the bioluminescence efficiency is concerned about here.

 

Fig. 7. The flash kinetics curves of the average bioluminescence intensity of dinoflagellates. A vertical error bar, standard error, is drawn at each data point. The error bar of Ceratium horridum does not exist because it was only measured once. The bioluminescence data of Dinophysis,Ceratium furca and Ceratium horridum are multiplied by 10 to display.

Download Full Size | PDF

Table 2. Bioluminescence efficiency of five dinoflagellate species.

View Table | View all tables in this article

As seen from the results of onboard laboratory measurements (Table 2), RNS has the highest bioluminescence efficiency. The maximum bioluminescence intensity of RNS is $\mathrm {3.7\times 10^{9}~photon~s^{-1}}$, which is 3 times higher that of Protoperidinium, 15 times higher than that of Dinophysis, and more than 70 times higher than that of Ceratium horridum and Ceratium furca. Although the flash kinetics of Gonyaulax ploygramma and Ceratium fusus are not detected in our experiment, according to reported results, they have a similar bioluminescence intensity order of $\mathrm {10^{8}~photon~s^{-1}}$ [22,23,37,40], which is much lower than that of RNS. The high bioluminescence efficiency from RNS is also necessary to retrieve RNS abundance using MBP.

According to Tables 1 and 2, the magnitude of MBP produced by approximately three Protoperidinium cells is considered to be close to that produced by one RNS cell. Approximately 14 Dinophysis cells could generate the same MBP as one RNS cell. More than 30 cells of Ceratium furca, Dinophysis, Gonyaulax ploygramma, Ceratium horridum or Ceratium fusus could produce a similar MBP as one RNS cell. Only when the abundance of other bioluminescent dinoflagellates is much higher than RNS abundance could the same MBP as that of RNS be produced in the field. For Ceratium fusus and Ceratium horridum, their abundance hardly exceeds dozens of times that of RNS in the YSBS. Thus, RNS can be considered as the most likely dinoflagellate leading to the high MBP rather than other dinoflagellates in the YSBS.

Through the above contents, the following points are also demonstrated. The error bar of each species displayed in Fig. 7 shows the range of variability, which indicates that bioluminescence intensity and bioluminescent flash kinetics can be aggregated across regions in the YSBS. Thus, using the data obtained from the large-scale field regions to establish the model of RNS abundance is reasonable. And the model should be consistent with the corresponding flash kinetic. However, the full and complete relationship between in situ MBP and bioluminescent flash kinetic of a certain dinoflagellate species, especially RNS, is exploring and requires more observations. This complex relationship is related to the difference of cell individuals revealed by the variability of flash kinetics. Even so, the model that is applicable to estimate RNS abundance using MBP, especially for relatively high abundance, reveals that RNS should have high bioluminescence intensity, which is confirmed by flash kinetic of RNS. Additionally, MPB actually represents bioluminescence intensity averaged in unit time per unit volume and flahs kinetic represents raw $\mathrm {60~Hz}$ bioluminescence intensity of a single living cell per unit time. Therefore, they are equivalent in theory and there is the internal consistency between the estimating model and flash kinetic, although their quantitative transformation relationship needs further study.

6. Conclusion

To further explore the relationship between bioluminescence and dinoflagellates, comprehensive in situ measurements combined with bioluminescence profiles, flash kinetics of dinoflagellate individuals and abundance of multiple dinoflagellates were conducted by a summer cruise in the YSBS. A total of eight bioluminescent dinoflagellates were found from seawater samples, but bioluminescence kinetics were obtained from five of the eight dinoflagellates, including RNS, Protoperidinium, Dinophysis, Ceratium horridum, and Ceratium furca. The comparison with the distributions of the MBP, total counted dinoflagellate abundance and proportion of the five dinoflagellate species indicated that the spatial distribution of bioluminescence varies with abundance and species of dinoflagellates in both the vertical and horizontal direction. And certain regions with high (or low) MBP and low (or high) abundance indicate that the level of MBP is related to dinoflagellate species rather than abundance. Bioluminescence flash kinetics are indeed related to dinoflagellate species according to laboratory measurements of individual dinoflagellate species. Combined with qualitative analysis of their distribution and quantitative analysis of flash kinetics, RNS has the highest bioluminescence efficiency $\mathrm {3.7\times 10^{9}~photon~s^{-1}}$ in the YSBS among the five dinoflagellate species , and its abundance is accurately estimated by bioluminescence with $\mathrm {R^{2}}$ of 0.82 and $\mathrm {RMSE}$ of 0.39 on a logarithmic scale . The MBP could be used as an optical proxy to estimate RNS abundance by the proposed method in this study. Considering the coexistence of other bioluminescent organisms and RNS in the field, the relative error of estimates is large when the RNS abundance is low. Even so, RNS abundance estimates by the MBP could not be underestimated at high abundance levels. In terms of early red tide warnings, there is a risk of RNS red tide when its abundance continues to increase or suddenly increases. Therefore, the MBP could be utilized in field monitoring and even be a predictor of red tide events in the YSBS.

In addition, the instruments for measuring the MBP or similar parameters are convenient to be deployed on multiple underwater platforms, so the proposed method has practicability in field monitoring. However, the quantitative parameters may need to be tuned based on in situ datasets for specific areas when referring to the proposed method, because the proposed method is based on a regional empirical statistical relationship, although its feasibility in the YSBS has been discussed in detail by bioluminescence flash kinetics. More field observations and detailed analysis are needed in future research. In addition, necessary conditions using the proposed method need to be emphasized. In brief, RNS should be the dominant dinoflagellate species and have the highest bioluminescence efficiency among the bioluminescent organisms in the field, which is the case in the YSBS. Thus, the proposed method is feasible in the YSBS and could be used as an early warning method for RNS red tide.