Open Access
Add to BibSonomy
Abstract
Secchi disk depth (Zsd), represents water transparency which is an intuitive indicator of water quality and can be used to derive inherent optical properties, chlorophyll-a concentrations, and primary productivity. In this study, the Zsd was derived from the Geostationary Ocean Color Imager (GOCI) data over the Bohai Sea (BHS) and the Yellow Sea (YS) using a regional tuned model. To validate the GOCI derived Zsd observations, in situ data, were collected for the BHS and YS regions. Results showed a good agreement between the GOCI derived Zsd observations and in situ measurements with a determination coefficient of 0.90, root mean square error of 2.17 m and mean absolute percent error of 24.56%. Results for diurnal variations showed an increasing trend of Zsd at the first and then decreasing, and all the maxima of Zsd in the central areas of the BHS and YS were found in the midday. For seasonal variations, higher values of Zsd, both in range and intensity, were observed in summer compared with those in winter. The reasons to explain the variations of Zsd have also been explored. Solar zenith angle (SOLZ) has an impact on the daily dynamics of Zsd, due to the influence of SOLZ on the attenuation of light radiation in water. The influence level of SOLZ on Zsd is largely determined by the water bodies’ composition. The significant seasonal variations are mainly controlled by the stability of the water column stratification, because it can lead to the sediment resuspension and influence the growth and distribution of phytoplankton. Runoff and sediment discharge are not the main factors that impact the seasonal dynamics of Zsd. Tidal currents and mean currents may have influences on the variations of Zsd. However, due to the lack of in situ measurements to support, further studies are still needed.
© 2018 Optical Society of America under the terms of the OSA Open Access Publishing Agreement
1. Introduction
Water transparency is an intuitive indicator of water quality which can be represented by Secchi disk depth (Zsd) [1]. Zsd is an important marine optical parameter and can be used to derive the inherent optical properties [2–5], the concentration of chlorophyll-a (Chl-a) [6] and primary productivity [7]. In general, Zsd negatively related with the concentrations of Chl-a, suspended particle matter (SPM), and chromophoric dissolved organic matter (CDOM), as these matters influence the absorption and scattering of the underwater light field. Moreover, the value of Zsd is also closely related to the distribution of the ambient light field.
Traditional measurements of Zsd depend on the cruise surveys and sampling sites, which are discrete and sparse, in general. Such measurements are helpful to understand temporal variations in Zsd, but these are not helpful for to perform spatial analysis due to spare sampling locations. Therefore, advancement in the satellite technology overcome this spatial limitation and can provide ocean color observations at high spatiotemporal coverage, which are ideal for examining the spatiotemporal variability of Zsd. Over the past decades, estimations of Zsd have been carried out using the polar orbit satellite sensors such as the Sea-viewing Wide Field-of-view Sensor (SeaWiFS) [8], Moderate Resolution Imaging Spectroradiometer (MODIS) [1], Medium Resolution Imaging Spectrometer (MERIS) [9], Landsat 8 Operational Land Imager (L8/OLI) [10]. However, these polar-orbiting satellites can provide observation only for one or two times a day, per satellite at mid-low latitudes [11], which are not sufficient to perform comprehensive analysis and to monitor variations in Zsd. On the other hand, the geostationary satellites can provide observations at a very high temporal resolution to monitor the daily dynamics of ocean waters. The Geostationary Ocean Color Imager (GOCI), the world’s first geostationary satellite ocean color sensors, was launched on June 27, 2010, and spatial coverage of about 2500 km × 2500 km (116.08 °E −143.92 °E, 24.75 °N-47.25 °N for central directions) covering the coastal area of Eastern China, the Korean peninsula, Japan and the adjacent shelves, and open oceans [11, 12]. In comparison to the polar-orbiting satellites, GOCI provides 8 hourly observations per day at 500 m spatial resolution [12], and these very temporal observations are very useful to study the diurnal variations of marine environments in sub mesoscale regions.
In previous studies, water transparency has been derived using a variety of remote sensing models, and most of them are empirical [13–15]. Such kind of models can be directly developed by establishing the empirical relationship between water transparency and spectral data or Chl-a concentrations. However, the empirical models have the following two limitations: (1) they require a large amount of match-up in situ measurements to calculate the models’ coefficients, and (2) the coefficients of the empirical models are location dependent, and may not be suitable for the same application to other waters [10]. To solve such limitations in empirically retrieving Zsd from remote sensing, the analytical models have been developed for retrieving Zsd [16] based on the classical underwater visibility theory [17, 18]. Nevertheless, it was found that the analytical Zsd model has shown low agreement with in situ measurements [16]. In this case, Lee et al. reported [19], the most likely reason for the low agreement, that the classical underwater visibility theory does not match the physical process of sighting a Secchi disk in water by human eye. Thus, a new underwater visibility theory was proposed and a new semi-analytical transparency model for the global ocean was developed by Lee et al. [19]. However, regarding the applicability of this new model, further studies are still required, especially for coastal waters.
The retrieval of Zsd in coastal waters is of great interest and has a practical significance due to the complexity of water optical properties and the importance of economic activities in coastal waters. The Bohai Sea (BHS) and the Yellow Sea (YS), the most turbid seas in the world [20], are located to the east of China. The seasonal variations of monsoon, mean currents, runoff, as well as the diurnal dynamics of tidal currents are expected to influence the water quality of the BHS and YS. However, there are limited studies for monitoring the variations in Zsd in the BHS and YS, especially based on high spatiotemporal satellite observations.
The objective of this study is to map the spatiotemporal distributions of Zsd in the BHS and YS and discuss its potential influencing factors. Lee et al.’s semi-analytical transparency model [19] was used in this study with modified retrievals of Kd, (the diffuse attenuation coefficient), which is the key parameter in Lee et al.’s Zsd model, for its applicability to the BHS and YS regions. The main data used in this study includes a large number of in situ measurements of Rrs (Remote sensing reflectance), Kd and Zsd, GOCI satellite data of Rrs to test the accuracy of GOCI retrieved Zsd. Besides, the data of ocean depth from National Oceanic and Atmospheric Administration (NOAA), GOCI solar zenith angle (SOLZ), in situ records of tidal level from Chinese seaports, runoff and sediment discharge of Yangze River as well as simulation data of wind fields and mean currents were taken as the auxiliary data to help us understand the variability of the Zsd in the BHS and YS.
2. Data and methods
2.1 Study area
The BHS, located in the east of China, is a large (77,000 km2) and shallow (~18 m on average) [1] semi-enclosed marginal sea. It consists of five parts: Liaodong Bay, Bohai Bay, Laizhou Bay, the central part of BHS and the Bohai Strait. In the BHS, the ability of the sea waters’ exchange is poor and the marine ecosystem is vulnerable. The YS, located to the east of BHS, is also a semi-enclosed sea with a mean depth of 44 m and area of 380,000 km2 [21] and contains the North Yellow Sea (NYS) and the South Yellow Sea (SYS). The BHS and YS are under the influences of many natural factors and hydrodynamic conditions. Tidal action can make difference between the BHS and YS’s coastal waters [22]. Some studies have shown that over the broad shallow China seas, the bottom friction dissipation caused by tidal currents can account for about 9% of the total global M2 (the main semidiurnal tide) tidal dissipation. In addition, the speed of tidal currents can reach up to 1 ms−1 [23–25]. Wind fields in the BHS and YS have obvious seasonal changes [26]. In winter, the study area is characterized by strong, northwesterly winds. In the BHS, winter monsoon with wind speed higher than 17 ms−1 occur 6.4 times on average each year, and for NYS and SYS the wind speeds are less than 8-9 ms−1. On the contrary, in summer, prevailing winds are southeasterly and weak with speeds 4-6 ms−1 in the BHS and YS [27–29]. In addition to the above natural factors, the BHS and YS also serve as primary receivers of many rivers. Among these rivers, the Yangtze River is the largest, with a mean water discharge 925 km3 yr−1 and a sediment load of 4.8 × 108 t yr−1 [11]. The direction and intensity of mean currents in the BHS and YS are consistent with the monsoon.
2.2 In situ data collection
In situ optical data used in this study were collected from two sources, named source A and source B. The source A data set includes the Kd and the Rrs (Remote sensing reflectance). This data set was collected in the turbid coastal waters of the China seas (the Pearl River Estuary and its adjacent waters (PRE), the Yangtze River Estuary and the East China Sea (ECS)) between July 2000 and February 2004, detailed information of this data set is given by Qiu et al. [20]. The source B data set includes the parameters of Kd, Rrs, and Zsd, which were collected from 4 oceanographic cruises in the BHS and YS between May 2014 and July 2016 (initiated by the National Foundation Commission). The summary of the data set used in this study is given in Table 1 and sampling stations of the Source B data set are shown in Fig. 1. The data sets from both sources A and B were used to validate the global, semi-analytical Kd algorithm developed by Lee et al. [30] and a regional tuned Kd algorithm, and only source B data set was used to validate the Zsd model which was based on this regional tuned Kd algorithm.
Table 1. Location and time of the cruise surveys to measure ocean properties
Fig. 1 (a) Map of the BHS, the north of the Yellow Sea (NYS) and the south of the Yellow Sea (SYS), the black rectangular boxes denote the studied regions which represent the coast and central areas of the BHS and YS. (b) Sampling sites for bio-optical properties of the water in BHS and YS, between May 2014 and July 2016, the red triangles represent the satellite-ground synchronous data sets based on this filed measurements and the red asterisks denote the tide stations of Yingkou, Jinzhou, Jingtang, Longkou, Yantai, and Lianyungang, respectively. (c) Map of Yangtze estuary. Based on the diffusion form of the waters in the river mouth, it is divided into three fan-shaped segments from the estuary: Z1, Z2, and Z3, respectively.
The measurements of Kd and Rrs in source A were conducted by PRR800 optical profiler (Biospherical Instruments Inc.) and the observations of Kd and Rrs in source B were conducted by Hyper-Profiler II (Satlantic Inc., Halifax, NS, Canada). These two instruments in operation processes are similar, but they have different band settings. The PRR800 optical profiler directly measures downwelling irradiance (Ed) and upwelling radiance (Lu) in 18 spectral channels centered at 340, 380, 395, 412, 443, 465, 490, 510, 520, 532, 555, 565, 625, 665, 683, 694, 710 and 765 nm, while the Hyper-Profiler II has more spectral channels from 350 to 804 nm with the sampling interval about 3 nm. In the measurements of Hyper-Profiler II, we interpolated the in situ spectral to obtain a new spectral data with the resolution of 1 nm.
At each station, the instrument was located about 10 m away from the operating boat, in order to avoid contact with the ship’s shadow. Then we kept the devices in the water for about 5 min in order to equilibrate the temperature of the instrument and the sea water. After that, we controlled the cable and sunk the device at about 0.5 m per second to record both the upcast and downcast data. At the same time, the dark offset measurements were recorded. Each cast would be repeated several times and the mean value of the measurements was used to ensure the data quality. In the above operation, data with tilt angles > 10 degrees were discarded. After correction of the measurement profile with the appropriate dark offset [20], the Eq. (1) was used to calculate Kd and the Eqs. (2) and (3) were used to calculate Rrs.
Where λ is the wave length; z represents the water depth; 0- represents the depth just beneath the water-air interface; and rrs is the spectral reflectance just beneath the water-air interface. The data from the first optical depth was used to determine Kd.For the collections of Zsd measurements in the source B, a Secchi disk, a white-and-black disk with a diameter of 30 cm [19], was used. Observers usually lower the Secchi disk on the ship, when the disk is no longer viewable, the corresponding depth is the Zsd.
2.3 Satellite data
GOCI, the world’s first geostationary ocean color satellite sensor, provides observations of the Korean Peninsula and its adjacent waters for six visible bands (412, 443, 490, 555, 660, and 680 nm) and two near-infrared bands (745 and 865 nm). It should be noted that GOCI has high signal-to-noise ratios, which may provide more accurate retrieval of Zsd and other ocean parameters [11].
In this study, the Rrs and SOLZ parameters were obtained from the GOCI Level-2P (L2P) product. The Level-2P (L2P) products were generated from the Level-1B products using GOCI Data Processing System (GDPS) released by the Korea Ocean Satellite Center (KOSC). The GOCI derived Zsd observations were validated against in situ Zsd measurements, and total 10 data points were matched by strictly synchronizing the GOCI and in situ Zsd measurements with time tolerance about ± 3 hours and without any spatial tolerance. Besides this, the GOCI data sets on 17, February 2016 with the low cloud coverage and the hourly data from December 2014 to November 2015 were also obtained.
The NOAA ocean depth data (http://www.ngdc.noaa.gov/mgg/global/etopo5.html), the field measurements of tide height obtained from China Maritime Service Network (http://ocean.cnss.com.cn/), the wind speeds from Climate Forecast System version 2 (CFSv2) (ftp://ftp.hycom.org/datasets/force/ncep_cfsv2/netcdf/), currents simulated from HYCOM (Hybrid Coordinate Ocean Model) (https://hycom.org/.) and the runoff records of Yangtze River from Yangtze River sediment bulletin (http://www.cjw.gov.cn/zwzc/bmgb/) were also included in this paper, in order to discuss the potential influence factors of the spatiotemporal variations of Zsd in the BHS and YS.
2.4 Retrieval of Zsd observations
Following the new underwater visibility theory, a semi-analytical model for estimating Zsd in global ocean has been developed and validated by Lee et al. [19]. In this study, this model (Eq. (4)) was implemented to generate GOCI Zsd observations for the BHS and YS. The Lee et al.’s model can be expressed as
where the wavelength of Rrs corresponding to the minimum Kd. It can be seen that Kd is the key parameter of Lee et al.’s model for estimating Zsd. Thus, to ensure the accuracy of Kd retrieval is of great importance [31]. In the Lee et al.’s model, two steps are used to obtain Kd: (1) the first step is to get absorption coefficient (a) and back-scattering coefficient (bb) from the Rrs based on the quai-analytical algorithm (QAA); (2) the second step is the estimation of Kd based on a and bb. Here, it should be noted that the calculations of Kd used in the semi-analytical Zsd model of Lee et al. [19] is for global ocean; thus, the model applicability in local regions, especially for coastal waters, need further validations.2.5 Model accuracy assessment
The precision tests of the models were performed with MATLAB software and evaluated by the coefficient of determination (R2), the root of mean square error (RMSE) and the mean absolute percent error (MAPE). These statistical indicators can be expressed as:
Where xi and yi indicate the in situ and derived Zsd for the ith sample, respectively; N represents the total number of samples. The coefficient of determination is the square of the correlation coefficient.3. Results
3.1 Validation of regional tuned Kd algorithm
In this study, the new, semi-analytical Kd global algorithm developed by Lee et al. [30] was validated against in situ Kd measurements collected for the source A and B in China seas (here, Kd(490) was selected) (Fig. 2). Results showed an obvious deviation at the high-value range of Kd(490) which indicates that the global Kd algorithm is not suitable for high turbid waters of China seas. Therefore, a regional tuned Kd algorithm developed by Mao et al. [32] for the China seas was used in this study and the description of the tuned Kd algorithm can be found in [32] and also briefly described here.
Fig. 2 Measured Kd(490) versus modeled Kd(490) developed by Lee et al. (2015). The red rectangle indicates the obvious deviation at the high-value range of Kd(490).
The semi-analytical Kd(490) algorithm developed by Mao et al. [32] is based on the idea presented in the work of Lee et al. [33], which can be expressed as:
Where θ0 is the SOLZ, a(490) and bb(490) are the absorption coefficient and back-scattering coefficient at the wavelength of 490 nm. The bb(490) was calculated using the nonlinear relationship between bb(490) and the irradiance reflectance at 490, 665 and 710 nm (Eqs. (8) and 9) [32], and this method is different from that used by Lee et al. [33]. The calculation of a(490) depends on the bb(490) and was carried out using Eq. (10). where f, Q, B, bbw(490), aw(665) and bbw(665) were taken as constants with the values of 0.305, 4, 1.13, 0.309, 3 × 10−4, 1.1 × 10−3, respectively. In order to apply Kd(490) algorithm to both clear and turbid waters, the regional tuned Kd algorithm is defined as a weighting algorithm as given in Eq. (11), where, Kd clear represents the global Kd algorithm developed by Lee et al. [30] and Kd turbid represents the Kd algorithm developed by Mao et al. [32], w1 and w2 are the weighting coefficients of Kd clear and Kd turbid, respectively, and calculated using ratio Rrs(683)/Rrs(490). Figure 3(a) showed that in the relative clear oceans, i.e. Kd(490) < 0.7 m−1 and Rrs(683)/Rrs(490) < 0.5, the scatter points is well distributed near 1:1 line. While in the high turbid waters, i.e. Kd(490) > 2 m−1 and Rrs(683)/Rrs(490) > 1.2, an obvious deviation was revealed. Therefore, when the values of Rrs(683)/Rrs(490) were less than 0.5, then the Kd clear algorithm was used (i.e., w1 = 1,w2 = 0 in Eq. (11)), and when the values of Rrs(683)/Rrs(490) were greater than 1.2, then the Kd turbid algorithm was used (i.e., w1 = 0, w2 = 1 in Eq. (11)). When the values of Rrs(683)/Rrs(490) were between 0.5 and 1.2, w1 and w2 were calculated using Eq. (12) and Eq. (13). Results (Fig. 3(b)) showed that the regional tuned Kd algorithm has a better performance for the high values of in situ Kd compared to the global Kd algorithm developed by Lee et al. [30] (Fig. 2). The values of R2, RMSE, and MAPE were 0.86, 0.80 m−1 and 36.69%, respectively. Table 2 shows the statistics of the results from Kd(490) algorithms before (Fig. 2) and after regional tuned (Fig. 3(b)).Fig. 3 (a) Kd(490) versus Rrs(683) / Rrs(490) . The area of the red lines is the transition between the clear or slightly turbid waters and the highly turbid waters. (b) Measured Kd(490) vs Modeled Kd(490) developed by combined model.
Table 2. The contrast between the accuracy of the Kd(490) algorithms before and after regional tuned
Based on the good linear relationship between Kd(490) and Kd from other spectral bands (443, 532, 555 and 665 nm) (Fig. 4, Table 3), Kd(443), Kd(532), Kd(555), Kd(665) were calculated using the regional tuned Kd(490) algorithm. Then, the spectral minimum value of Kd was calculated and used in Eq. (4) to calculate Zsd. As GOCI data lack of Rrs(709) for calculating the Kd(490) in high turbid waters, therefore, a multiple regression method [34] was used to derive Rrs(709) from the other existing spectral bands. Initially, the regional tuned Kd(490) algorithm was used to calculate Zsd based on the in situ measurements and derived Zsd was validated against in situ Zsd measurements (Fig. 5(a)). Results showed that the values of R2, RMSE, and MAPE for in situ Zsd versus modeled Zsd were 0.89, 1.67 m and 29.26%, respectively. Later, the regional tuned Kd(490) algorithm was applied to the GOCI data to derive Zsd observations and validated against in situ Zsd measurements (Fig. 5(b)). Those of satellite-ground synchronous data sets versus modeled Zsd were 0.90, 2.17 m and 24.56%, respectively (Fig. 5). These results provide us the confidence to estimate spatial and temporal variations of Zsd in the BHS and YS.
Table 3. The linear regression results between Kd(490) and other bands of Kd.
Fig. 5 (a) Modeled Zsd (based on in situ Kd) versus in situ Zsd; (b) GOCI derived Zsd versus in situ Zsd.
3.2 Diurnal variations of GOCI derived Zsd in BHS and YS
Eight hourly images of GOCI data were obtained for February 17, 2016, and diurnal variations of Zsd in the BHS and YS were calculated using the Zsd model of Lee et al. [19] and the regional tuned Kd algorithm of Mao et al. [32] (Fig. 6). In general, Zsd is low in the whole the BHS region, while in the YS region, clear gradients can be found. In Fig. 6, the color variations from yellow to blue representing low to high transparency which were revealed from the coast to the offshore waters. Temporal variations showed that the area of higher Zsd gradually expanded to the northern waters before the midday (Fig. 6(a) to 6(d)), while the area of higher Zsd gradually shrunk to the southern regions after the midday (Fig. 6(e) to 6(h)).
Fig. 6 Hourly spatiotemporal variations of GOIC derived Zsd in the BHS and YS on February 17, 2016. Where white areas represent cloud covers.
In order to see these diurnal variations more clearly, three rectangular regions A, B, D (Fig. 1(a)), and six tidal stations Yingkou (YK), Jinzhou (JZ), Jingtang (JT), Longkou (LK), Yantai (YT) and Lianyungang (LYG) (Fig. 1(b)) were selected to represent the central and coast areas of BHS and YS. Figure 7 shows the diurnal variations of GOCI derived Zsd in the central areas of the BHS, NYS, and SYS on September 2015, as most of the observations available during this month. In Fig. 7, BHS has a lower value of Zsd compared to the NYS and SYS at the same time. Also, the values of Zsd in the central parts of BHS, NYS, and SYS were increased before the midday (the time corresponding to the minimum of SOLZ as the sun is directly above at this time) and then gradually decreased after the midday.
Fig. 7 Diurnal variations in the average values of GOCI derived Zsd, represented by red lines with red points, in the central areas of BHS (Box A), NYS (Box B) and SYS (Box D) as shown in Fig. 1(a) as well as the corresponding SOLZ, represented by blue lines with blue points, observations in September 2015.
The diurnal variations of SOLZ were also analyzed to explore the relationship between SOLZ and Zsd (Fig. 7). Figure 7 showed an inverse relationship between Zsd and SOLZ and the values of the correlation coefficient between Zsd and SOLZ in the central parts of BHS, NYS, and SYS were −0.60, −0.76, −0.77, respectively. By contrast, the Zsd daily changes in the coastal waters were more irregular. Figure 8 showed the average values of GOCI derived Zsd near the tidal stations within a spatial range of ± 5 km in the BHS and YS in September 2015. In most coastal waters, the Zsd increased at first and then decreased, the maxima of Zsd appeared at different times. At YK and JZ tidal stations, the values of the correlation coefficient between Zsd and SOLZ were −0.75, −0.77, respectively; however, at JT, LK, YT and LYG stations, the values of correlation coefficient were less than −0.60. Whereas, the poor relationship between Zsd and the tidal level was observed as the values of correlation coefficient were less than 0.45.
Fig. 8 Diurnal variations in the average values of GOCI derived Zsd, represented by red lines with red spots, and SOLZ, represented by blue lines with blue spots, near the tidal stations at a spatial range of ± 5 km in the BHS and YS and in September 2015 as well as the corresponding field tidal level records, represented by black lines with black spots. Where R1 represents the correlation between Zsd and SOLZ and R2 represents the correlation between Zsd and tidal level.
3.3 Monthly variations of GOCI derived Zsd in BHS and YS
Monthly mean Zsd observations were obtained from GOCI hourly Zsd observations. In detail, the GOCI hourly images of Zsd in one month were added at first and then divided by the total number of the valid data in that month. Figure 9 showed the monthly mean Zsd observations, which were obtained from the GOCI hourly Zsd observations, from December 2014 to November 2015 in the BHS and YS and distinct spatiotemporal dynamics of Zsd were revealed. In the summer months (June, July, and August), the Zsd was generally high as the values were greater than 14 m extended to the central part of BHS, while the values were less than 3 m in the Liaodong Bay, Bohai Bay and Laizhou Bay regions of BHS. It should be noted that low Zsd also existed in the coastal waters of YS, but only extended to 122.6 °E from the coast of China to the offshore regions. In the winter months (December, January, and February), the area of Zsd values greater than 14 m was shrunk to the SYS and almost the whole area of BHS had a relative low Zsd. It was found that the values of Zsd less than 3 m were extended to 126 °E in the YS from the coast of China. In the spring (March, April, and May) and the autumn months (September, October, and November), the distribution and the variations of Zsd were the transitions between summer and winter months.
Fig. 9 Monthly average observations of GOCI derived Zsd in the BHS and YS from December 2014 to November 2015.
To further explore the potential impact factors causing variations in Zsd, the monthly mean Zsd observation were analyzed with respect to wind speed observations from December 2014 to November 2015 (Fig. 10) for the regions A, B, C, D, E, F and G as marked in the Fig. 1. Results showed a negative relationship between Zsd and wind speed observations at both the coast and the central parts of the study area with R values greater than 0.70 and the p values < 0.05. Results also showed that the Zsd decreased with increased of wind speeds in the winter months, while opposite trend was observed in the summer months.
Fig. 10 The monthly mean GOCI derived Zsd observations, represented by blue lines with blue points, in the central and coastal parts of BHS and YS (Box A, B, C, D, E, F, G, as shown in Fig. 1) versus the monthly average values of the wind speed, represented by black lines with black points, derived from CFSv2 data from December, 2014 to November, 2015. R is the correlation coefficient between Zsd and wind speed, P is the P-value.
4. Discussion
4.1 Controlling factors of the Zsd diurnal dynamics in BHS and YS
Many factors can affect diurnal variations of the Zsd in the BHS and YS including the SOLZ and tidal currents. The daily variations of SOLZ are regular, i.e., when the time close to the midday, the values of SOLZ decreases, and afternoon, the values of SOLZ gradually increases. Verschuur [35] found the influence of SOLZ on the Zsd based on the measurements of nearly three years. When the angle of artificial light in the evening became same as the SOLZ in the daytime, the observations of Zsd at night time were close to the measurements during the daytime. The reason of this phenomenon is that Zsd has a strong negative relationship with Kd (Eq. (4)), which is one of the important factors and can be used to characterize the attenuation of light in water, and also, it is an apparent optical parameter which can be affected by the surrounding light field). Kirk [36] found a positive relationship between Kd and SOLZ based on the Monte Carlo calculation, and this relationship is particularly an evident in the clear waters.
In this study, the values of Kd was varied significantly with the SOLZ in the central parts of the BHS, NYS, and SYS, where the water bodies are clear, and as a result, this led to the variations of Zsd. This indicated that the diurnal variations of Zsd in the central parts of BHS and YS were mainly controlled by SOLZ as good relationship between Zsd and SOLZ can be seen in Fig. 7. In contrast to the central parts of the BHS and YS, the relationship between Zsd and SOLZ in the coastal waters were different from each other as shown in Fig. 8. These results suggest that the influences of the SOLZ are different for the different coastal regions. This might be due to the complex water bodies’ composition and their optical characteristics in the coastal regions than the central waters. As a result, the sensitivity of Kd to the SOLZ has changed. Especially, the more obvious of the scattering characteristic of coastal waters, the less sensitive of Kd to the SOLZ [36], and therefore reduce the correlation between Zsd and SOLZ.
Different from the mechanism of the Zsd’s variations caused by SOLZ, tidal current mainly influences Zsd through changing the composition of the water bodies. As one of the most important hydrodynamic conditions in coastal regions, the most significant feature of tidal currents is the transport of material. During the period of the flood tide, the onshore flows would bring clear outer shelf waters into the coastal regions, while during the period of the ebb tide, the coastal waters would be under the influence of dirty land-based pollutants that brought by offshore currents. However, in our study, there is a lack of covariance between the Zsd and the tidal level records as shown in Fig. 8. In this case, we must state that this does not mean that the tidal currents have no effect on Zsd. As along with the above processes, the tidal currents would be rubbed with the bottom to induce the sediment resuspension and then affect the values of Zsd. The greater the velocity of currents is, the more easily the sediment to be resuspended. However, due to the lack of the field measurements of tidal currents’ velocity, we cannot examine the potential impacts of tidal currents at the current state, and further studies are still needed.
4.2 Relationships between monthly Zsd variations and environment factors
Monthly changes of Zsd are relevant to water column stratification which represents the degree of the seawater’s turbulence. Water column stratification is mainly affected by wind-wave, water temperature and so on. Generally, weak water column stratification corresponding to the strong seawater’s turbulence. When the strong seawater’s turbulence happens, it can cause the bottom sediment to be suspended to the upper waters and then reduce the Zsd. In addition, the strong seawater’s turbulence can also induce the underlying nutrients into the surface layer to facilitate the growth and distribution of phytoplankton, and as a result, the concentration of chlorophyll can also increase in the upper waters and lead to the low values of Zsd. In contrast, strong water column stratification corresponding to the weak seawater’s turbulence, it cannot only lead to the deposition of suspended sediments in the water but also hinder the underlying nutrients into the surface layer. As a result, in the upper waters, the content of sediments decreased and the growth and reproduction of phytoplankton reduced which lead to the high values of Zsd. Figure 11 shows the degree of the turbulence of the BHS and YS from December 2014 to November 2015 using the Simpson and Hunter index (SH) [37], which can be calculated as follows:
where, H denotes the water depth and was obtained from National Oceanic and Atmospheric Administration (NOAA) (http://www.ngdc.noaa.gov/mgg/global/etopo5.html); U denotes the depth average of the monthly mean currents, corresponding data was drawn from the official website of HYCOM (https://hycom.org/). The higher values of SH, the worse turbulence of the water bodies, and vice versa.Figure 11 revealed that the values of SH and its monthly distribution were similar to that of Zsd. In the summer months, the values of SH were obviously higher than that in the winter months, as for spring and autumn months, the values of SH were in transition. The reason for this phenomenon is that the strong solar radiation caused the upper waters’ temperature to be higher than that of the bottom layer in the summer months. As a result, the density of the upper body was much smaller than that of the bottom. Also, in summer, the wind fields were the weakest which was bad for wave growth and rejected water mixing. As a result, the stability of the water column stratification was further strengthened and the turbulent of the waters was hard to occur, which led to the highest values of Zsd in the whole year. In the winter months, the solar radiation was weak, thus led to the surface layer temperature of the study area was low and the difference of the density between the upper and lower layers was little. In addition, in winter, the wind fields were strongest in the whole year, which generated strong waves and enhanced water mixing. Therefore, the water column stratification was broken, the water turbulence was strengthened and the values of Zsd were the lowest in the whole year.
Also, as the runoff is able to transport a large amount of sediments to the coastal waters every year, therefore, it is necessary to discuss the influence of river runoff and sediment discharge on Zsd. Here, the Yangtze River, China’s largest river, was taken as a case to study the potential impacts of river discharge on Zsd during the period from December 2014 to November 2015. As Datong hydrological gauging station is the last hydrological gauging station of Yangtze River, the runoff and sediment discharge of this gauging station can stand for the distributions of water and sediment at the estuary of the Yangtze River. The records of runoff and sediment discharge of Datong hydrological gauging station in the wet season (from May to October) were accounted for 65% and 69%, respectively, of the whole year (see the Yangtze River sediment bulletin for more information, http://www.cjw.gov.cn/zwzc/bmgb/). These records indicated that at the mouth of the Yangtze river, the sediment loading per unit was higher in the wet season and lower in the dry season (from November to April), and the runoff, sediment discharge as well as sediment loading per unit have the consistent monthly variation trend. However, as shown in Fig. 9, in the wet season, the area of high Zsd around the Yangtze estuary was larger than that in the dry season. This situation evidently violates the fact that the negative relationship between the amount of sediment and the values of Zsd. To further illustrate this phenomenon, based on the diffusion form of the waters in the river mouth, three fan-shaped areas (shown in Fig. 1, named Z1, Z2, and Z3, respectively) around the Yangtze estuary were selected for discussion. Figure 12(a), 12(b), and 12(c) showed the monthly variations’ comparison between the records of the runoff in Datong hydrological gauging station and the GOCI-retrieved mean Zsd in the Z1, Z2, and Z3, respectively, from December 2014 to November 2015. As shown in Fig. 12, the monthly Yangtze River discharge in wet season was significantly higher than that in the dry season. Also, the correlation coefficients between Zsd and the runoff in Z1, Z2 and Z3 gradually increased (0.3, 0.85 and 0.92 for Z1, Z2, and Z3, respectively). These results indicated that runoff was not the main factor that caused the monthly dynamics of Zsd in the coastal waters. The reason for this phenomenon was probably due to the fact that at the mouth of the river, the sediments transport from the river to the sea will be quickly deposited [38, 39]. What’s more, in the Yangtze estuary, the deposition of the land source sediments has significant seasonal variations. That is, in summer, the sediments’ deposition effect was strong, and as a result, it was difficult for the bottom sediments to suspend to the upper layer. However, for the winter months, the situation was the opposite [38].
Fig. 12 Comparison of the monthly variations (from December 2014 to November 2015) between the records of the runoff in Datong hydrological gauging station and the GOCI-retrieved mean Zsd in the Z1, Z2, and Z3 (as marked in Fig. .1), respectively.
Apart from the influencing factors mentioned above, the monthly changes of Zsd may also under the influence of mean currents. It was because mean currents can transport the sediment and other materials. Some studies pointed out that [27, 40], in winter, under the effect of winter monsoon, the alongshore currents will be stronger and southerly. Therefore, a large amount of suspended sediments from the Yellow River, Yangtze River, and Subei coastal water could be easily carried to the south and then reduce the values of Zsd. In contrast, in summer, the alongshore currents will be weaker and northerly under the effects of the summer monsoon, thus the transport capacity of sediment is greatly weakened and lead to the values of Zsd increase.
5. Conclusion
In this study, the GOCI data was quantitatively used to map the diurnal and monthly distributions of Zsd in BHS and YS. The semi-analytical Zsd model developed by Lee et al. [19] was adjusted by developing a regional tuned Kd model for its applicability to BHS and YS. The GOCI derived Zsd observations based on the regional tuned Kd model was validated against in situ Zsd measurements. We found that the GOCI observations, in conjunction with the adjusted Zsd retrieval model, can be used to quantify Zsd distributions and variations for the BHS and YS. The water transparency in the BHS and YS has obvious diurnal and monthly variations. Daily dynamics of Zsd were showed increased at first and then decreased. Besides, all the maximum of Zsd in the central areas of the BHS and YS were found in the midday. This is due to that the daily changes of Zsd can be impacted by SOLZ, and the degree of influence is related to the compositions of water bodies. Obvious monthly changes of Zsd was reflected high in summer but low in winter. This was mainly attributed to monthly changes of Zsd under the influence of sediment resuspension and the growth and distribution of phytoplankton in the upper waters, as these two processes were determined by the stability of water column stratification. River runoff and sediment discharge were not the main factors that affect the monthly variability in Zsd distributions.
Note that the dynamics of Zsd may also be impacted by tidal currents and mean currents. However, due to the lack of corresponding field measurements to support, further studies are still needed.
Funding
National Key Research and Development Program of China (No. 2016YFC1400901 and 2016YFC1400904); National Natural Science Foundation of China (No. 41576172 and 41506200); Provincial Natural Science Foundation of Jiangsu in China (No. BK20151526, BK2015914, BK20161532); National Program on Global Change and Air-sea Interaction (No. GASI-03-03-01-01); Priority Academic Program Development of Jiangsu Higher Education Institutions (PAPD).
Acknowledgment
We acknowledge the captains, officers, and crews of R/V Dongfanghong 2 for providing excellent assistance during field sampling and measurements. Special thanks to anonymous reviewers for their constructive comments towards improving this manuscript.
References and links
1. S. Shang, Z. Lee, L. Shi, G. Lin, G. Wei, and X. Li, “Changes in water clarity of the Bohai Sea: Observations from MODIS,” Remote Sens. Environ. 186, 22–31 (2016). [CrossRef]
2. Z. Chen, F. E. Muller-Karger, and C. Hu, “Remote sensing of water clarity in Tampa Bay,” Remote Sens. Environ. 109(2), 249–259 (2007). [CrossRef]
3. S. Kratzer, B. Håkansson, and C. Sahlin, “Assessing Secchi and Photic Zone Depth in the Baltic Sea from Satellite Data,” Ambio 32(8), 577–585 (2003). [CrossRef] [PubMed]
4. I. M. Levin and T. M. Radomyslskaya, “Estimate of water inherent optical properties from Secchi depth,” Izv Atmos. Ocean Phy. 48(2), 214–221 (2012). [CrossRef]
5. S. H. Son, J. Campbell, M. Dowell, S. Yoo, and J. Noh, “Primary production in the Yellow Sea determined by ocean color remote sensing,” Mar. Ecol. Prog. Ser. 303, 91–103 (2005). [CrossRef]
6. S. Kosten, M. Vernooij, E. H. Van Nes, M. D. Á. G. Sagrario, J. G. P. W. Clevers, and M. Scheffer, “Bimodal transparency as an indicator for alternative states in South American lakes,” Freshwater Biol 57(6), 1191–1201 (2012). [CrossRef]
7. P. G. Falkowski and C. Wilson, “Phytoplankton productivity in the North Pacific ocean since 1900 and implications for absorption of anthropogenic CO 2,” Nature 358(6389), 741–743 (1992). [CrossRef]
8. X. He, D. Pan, and Z. Mao, “The study on the inversing model of water transparency using the SeaWiFS data,” Acta Oceanol. Sin. 26, 55 (2004).
9. H. Taheri Shahraini, H. Sharifi, and M. Sanaeifar, “Development of clarity model for Caspian Sea using MERIS data,” in SPIE Remote Sensing(2011), pp. -.
10. Z. Lee, S. Shang, L. Qi, J. Yan, and G. Lin, “A semi-analytical scheme to estimate Secchi-disk depth from Landsat-8 measurements,” Remote Sens. Environ. 177, 101–106 (2016). [CrossRef]
11. X. He, Y. Bai, D. Pan, N. Huang, X. Dong, J. Chen, C. T. A. Chen, and Q. Cui, “Using geostationary satellite ocean color data to map the diurnal dynamics of suspended particulate matter in coastal waters,” Remote Sens. Environ. 133, 225–239 (2013). [CrossRef]
12. J. H. Ryu, J. K. Choi, J. Eom, and J. H. Ahn, “Temporal variation in Korean coastal waters using Geostationary Ocean Color Imager,” J. Coast. Res. SI64, 1731–1735 (2011).
13. C. E. Binding, J. H. Jerome, R. P. Bukata, and W. G. Booty, “Trends in Water Clarity of the Lower Great Lakes from Remotely Sensed Aquatic Color,” J. Great Lakes Res. 33(4), 828–841 (2007). [CrossRef]
14. S. Koponen, J. Pulliainen, K. Kallio, and M. Hallikainen, “Lake water quality classification with airborne hyperspectral spectrometer and simulated MERIS data,” Remote Sens. Environ. 79(1), 51–59 (2002). [CrossRef]
15. H. T. Shahraini, H. Sharifi, and M. Sanaeifar, “Development of clarity model for Caspian Sea using MERIS data,” Remote Sensing of the Ocean, Sea Ice, Coastal Waters, and Large Water Regions 2011. Vol. 8175. International Society for Optics and Photonics, (2011).
16. M. Doron, M. Babin, O. Hembise, A. Mangin, and P. Garnesson, “Ocean transparency from space: Validation of algorithms estimating Secchi depth using MERIS, MODIS and SeaWiFS data,” Remote Sens. Environ. 115(12), 2986–3001 (2011). [CrossRef]
17. S. Q. Duntley, “The Visibility of Submerged Objects” Visibility Lab., Mass. Inst. Tech. (pp. 74) (San Diego), (1952).
18. R. W. Preisendorfer, “Secchi Disk Science: Visual Optics of Natural Waters,” Limnol. Oceanogr. 31(5), 909–926 (1986). [CrossRef]
19. Z. P. Lee, S. Shang, C. Hu, K. Du, A. Weidemann, W. Hou, J. Lin, and G. Lin, “Secchi disk depth: A new theory and mechanistic model for underwater visibility,” Remote Sens. Environ. 169, 139–149 (2015). [CrossRef]
20. Z. Qiu, T. Wu, and Y. Su, “Retrieval of diffuse attenuation coefficient in the China seas from surface reflectance,” Opt. Express 21(13), 15287–15297 (2013). [CrossRef] [PubMed]
21. C. Hu, Z. Chen, T. D. Clayton, P. Swarzenski, J. C. Brock, and F. E. Muller-Karger, “Assessment of estuarine water-quality indicators using MODIS medium-resolution bands: Initial results from Tampa Bay, FL,” Remote Sens. Environ. 93(3), 423–441 (2004). [CrossRef]
22. Z. Hu, D. P. Wang, D. Pan, X. He, Y. Miyazawa, Y. Bai, D. Wang, and F. Gong, “Mapping surface tidal currents and Changjiang plume in the East China Sea from Geostationary Ocean Color Imager,” J. Geophys. Res-Oceans 121, 1563 (2016).
23. F. Lefèvre, C. L. Provost, and F. H. Lyard, “How can we improve a global ocean tide model at a regional scale? A test on the Yellow Sea and the East China Sea,” J. Geophys. Res. Oceans 105(C4), 8707–8725 (2000). [CrossRef]
24. H. J. Lie, S. Lee, and C. H. Cho, “Computation methods of major tidal currents from satellite-trackeddrifter positions, with application to the Yellow and East China Seas,” J. Geophys Res: Oceans 107, 22 (2002).
25. C. L. Provost and F. Lyard, “Energetics of the M 2 barotropic ocean tides: an estimate of bottom friction dissipation from a hydrodynamic model,” Prog. Oceanogr. 40(1-4), 37–52 (1997). [CrossRef]
26. H. J. Lee and S. Y. Chao, “A climatological description of circulation in and around the East China Sea,” Deep Sea Res. Part II Top. Stud. Oceanogr. 50(6-7), 1065–1084 (2003). [CrossRef]
27. L. X. Dong, W. B. Guan, Q. Chen, X. H. Li, X. H. Liu, and X. M. Zeng, “Sediment transport in the Yellow Sea and East China Sea,” Estuar. Coast. Shelf Sci. 93(3), 248–258 (2011). [CrossRef]
28. Z. Qiu, C. Xiao, W. Perrie, D. Sun, S. Wang, H. Shen, D. Yang, and Y. He, “Using Landsat 8 data to estimate suspended particulate matter in the Yellow River estuary,” J. Geophys. Res. Oceans 122, 276 (2017).
29. Y. Yu, H. Wang, X. Shi, X. Ran, T. Cui, S. Qiao, and Y. Liu, “New discharge regime of the Huanghe (Yellow River): Causes and implications,” Cont. Shelf Res. 69, 62–72 (2013). [CrossRef]
30. Z. Lee, C. Hu, S. Shang, K. Du, M. Lewis, R. Arnone, and R. Brewin, “Penetration of UV‐visible solar radiation in the global oceans: Insights from ocean color remote sensing,” J. Geophys. Res. Oceans 118(9), 4241–4255 (2013). [CrossRef]
31. T. Rodrigues, E. Alcântara, F. Watanabe, and N. Imai, “Retrieval of Secchi disk depth from a reservoir using a semi-analytical scheme,” Remote Sens. Environ. 198, 213–228 (2017). [CrossRef]
32. Y. Mao, Z. Qiu, D. Sun, S. Wang, Y. Lu, C. Wu, X. Yue, and Z. Ye, “A Novel Remote Sensing Algorithm for Estimating Diffuse Attenuation Coefficient in the BohaiSea and Yellow Sea,” Guangxi Science 23, 513–519 (2016).
33. Z. Lee, K. L. Carder, and R. A. Arnone, “Deriving inherent optical properties from water color: a multiband quasi-analytical algorithm for optically deep waters,” Appl. Opt. 41(27), 5755–5772 (2002). [CrossRef] [PubMed]
34. Z. Lee, S. Shang, C. Hu, and G. Zibordi, “Spectral interdependence of remote-sensing reflectance and its implications on the design of ocean color satellite sensors,” Appl. Opt. 53(15), 3301–3310 (2014). [CrossRef] [PubMed]
35. G. Verschuur, “Transparency Measurements in Garner Lake, Tennessee: The Relationship between Secchi Depth and Solar Altitude and a Suggestion for Normalization of Secchi Depth Data,” Lake Reserv. Manage. 13(2), 142–153 (1997). [CrossRef]
36. J. T. O. Kirk, “Dependence of Relationship Between Inherent and Apparent Optical Properties of Water on Solar Altitude,” Limnol. Oceanogr. 29(2), 350–356 (1984). [CrossRef]
37. J. H. Simpson and J. R. Hunter, “Fronts in the Irish Sea,” Nature 250(5465), 404–406 (1974). [CrossRef]
38. G. Guo, Z. Yang, D. Fan, and Y. Pan, “Seasonal Sedimentary Effect on the Changjiang Estuary Mud Area,” Acta Geogr. Sin. 58, 591–597 (2003).
39. J. Zhang, W. Huang, and M. Liu, “Distribution of Suspended Materials and its Seasonal Change in the Huanghei River Estuary as well as the nearby Sea area,” J. Ocean Univ. China 15, 96–104 (1985).
40. H. Xiao, H. Wang, N. Bi, X. Wu, A. Wang, and Y. Zhang, “Seasonal Variation of Suspended Sediment in the Bohai and Yellow Sea and the Pathway of Sediment Transport,” Haiyang Dizhi Yu Disiji Dizhi 35, 11–21 (2015).
