A Three-Dimensional Spatial Interpolation Method and Its Application to the Analysis of Oxygen Deficit in the Bohai Sea in Summer

Abstract

Dissolved oxygen (DO) plays a pivotal role in sustaining marine ecosystems. The Bohai Sea in China is a semi-enclosed sea, and oxygen-deficit events occur from time to time due to human activities. At present, there is a notable absence of any convenient and precise method for obtaining three-dimensional spatial data on DO, and the exploration of the physical mechanisms influencing oxygen deficit remains incomplete. This investigation uses the linear radial basis function (RBF-Linear) fitting method to conduct three-dimensional spatial interpolation for DO, which demonstrates minimal inaccuracy. Then, the RBF-Linear fitting method is utilized to collect a comprehensive three-dimensional spatial dataset encompassing temperature, salinity, and DO in the Bohai Sea in August from 2016 to 2018. The results indicate discernible interannual variations in the extent, area, and distribution of oxygen deficiency during summer in the Bohai Sea. Mechanism analysis reveals that intense precipitation episodes and an increase in wind stress curl exacerbate oxygen depletion. Additionally, the degree, location, and area of the two oxygen-deficit cores (off the Yellow River Estuary and off the Qinhuangdao) in the Bohai Sea are influenced by several factors, including current velocity, direction, local circulation position, and net horizontal transport rate. Furthermore, the study suggests that oxygen deficiency in the Bohai Sea region is currently in its early stages, with a limited degree of injury and a restricted range of influence. The use of a three-dimensional spatial interpolation method to create a complete DO field in three-dimensional space simplifies the research challenges associated with marine oxygen deficit. Moreover, this study holds particular significance for guiding the development of marine fisheries.

1. Introduction

Dissolved oxygen (DO) is a fundamental biochemical parameter in the ocean and a critical indicator of seawater quality. The occurrence of oxygen deficiency has dire consequences for marine organisms, including a decrease in the diversity, abundance, and biomass of benthic communities [1]. In addition, it results in the production of greenhouse gases (such as nitrous oxide and methane) and toxic compounds (such as hydrogen sulfide) [2,3]. Moreover, low-oxygen areas are frequently accompanied by the occurrence of seawater acidification [4,5], which is expected to pose a threat to calcified structures and lead to a significant decline in shellfish fishery [6]. In recent years, the number of coastal regions with oxygen deficiency has surged due to the ocean eutrophication caused by human activity [7]. At present, there are over 500 coastal areas worldwide with DO concentrations lower than 2 mg/L, accompanied by an increase in the area (volume) and the degree of hypoxia [7].

In the coastal regions of China, hypoxia and acidification have occurred in the estuaries of large rivers such as the Changjiang River [8,9,10,11,12] and the Pearl River [13,14] since the late 1950s, while the Bohai Sea (Figure 1) has newly developed an oxygen deficit in the past 20 years. In 2012, the oxygen deficit in the Bohai Sea began to receive attention when it was first reported by Zhai et al. [15]. Later, long-term observations revealed a substantially lower summer DO concentration near the seafloor in 2006–2018 than in 1978–2005 [16]. The oxygen-deficient area is primarily distributed in a “V”-shaped trough region in the central Bohai Sea and tends to generate two oxygen-deficit centers located in the offshore regions of the Yellow River estuary (YRE) and Qinhuangdao (QHD) [17]. Summer stratification in the Bohai Sea is a prerequisite for the formation of oxygen deficit in the bottom layer. Additionally, the intensity, starting time, and duration of stratification have obvious effects on the oxygen deficit in the bottom water [17,18]. Meanwhile, circulation and lateral transport modulate the bottom oxygen-deficit area in the central Bohai Sea [17,18,19]. In addition, pelagic respiration linked to the degradation of fresh organic matter prevails over sediment respiration in contributing to the oxygen depletion and acidification observed in this shallow water body [20,21].

At present, the spatial distribution of DO is mainly obtained by using scatter observation data to obtain the spatial distribution over part of a specific sea area [19,21] or by using marine ecosystem dynamics model to obtain the overall spatial distribution [18,22,23]. However, the former approach is limited in its capacity to comprehensively examine the complete spatial distribution of DO, while the latter method necessitates substantial computational resources and storage capacity. Hence, it is imperative to select a rational, precise, and effective method for 3D spatial interpolation. Furthermore, current research on hypoxia primarily focuses on the impact of water stratification and biochemical processes on the extent of hypoxia within a specific area [24,25,26]. This emphasis is particularly evident in studies conducted in the Bohai Sea region [16,20,22,27]. Nevertheless, there exists a paucity of analysis regarding other physical mechanisms and the spatial distribution and impact of oxygen-deficient areas. Consequently, further attention and research are warranted in addressing these questions.

Therefore, this study first compares a variety of spatial interpolation methods and selects a method that has a small error and is suitable for temperature, salinity, and DO spatial interpolation, namely, the linear radial basis function (RBF-Linear) fitting method. This method is widely used for interpolation of dispersed data due to its unconstrained size and high precision [28,29,30,31]. Second, this method is used to obtain the complete 3D spatial fields of temperature, salinity, and DO in the Bohai Sea in August from 2016 to 2018, and the spatial and temporal distributions of these elements are analyzed in detail. Then, the effects of air–sea interaction, ocean circulation, and horizontal transport rate on the bottom oxygen deficit in the Bohai Sea are analyzed. Finally, the factors affecting the area and volume of the oxygen-deficit region are discussed, which can provide some significant guidance for the development of marine fisheries.

2. Data and Methods

2.1. In Situ Observation and Reanalysis Data

The Bohai Sea is a semi-enclosed, shallow sea with an average depth of approximately 18 m [32] and an area of 77,000 square kilometers. The Bohai Sea is composed of five sub-regions: Bohai Bay (BHB), Liaodong Bay (LDB), Laizhou Bay (LZB), the central Bohai Sea (CBS), and the area off Qin Huangdao (QHD). Furthermore, the Bohai Sea is fed by several major rivers, including the Yellow River, Luan River, Hai River, and Liao River, as depicted in Figure 1a. According to previous studies [6,20], the level of oxygen deficit in the Bohai Sea is generally the most serious in August; accordingly, the temperature, salinity, and dissolved oxygen data in August are extracted from the data provided by the National Marine Environmental Monitoring from 2016 to 2018. The positions of the stations are shown in Figure 1a. Additionally, the conventional section observation data of the former State Oceanic Administration of China were used to obtain temperature data for the Yellow River estuary–Liaodong Bay section (mid-transect) in August from 2016 to 2018, as illustrated in Figure 1b. The specific sampling dates of Sta. A and Sta. B are shown in

Table 1. Sampling times of Sta. A and Sta. B.

Station	Year
	2016	2017	2018
Sta. A	14 August	10 August	14 August
Sta. B	5 August	10 August	14 August

. The result of the interpolation in the following chapters was to combine the two datasets and finally obtain the three-dimensional spatial dataset of the whole field. The temperature and salinity datasets were obtained using a conductivity–temperature–depth (CTD) sensor, while DO samples were collected, fixed, and titrated on board following the Winkler procedure.

Following the integration of the two datasets, Figure 2 illustrates the spatial distribution of observational data in the Bohai Sea in August from 2016 to 2018. The quantities of temperature, salinity, and dissolved oxygen (DO) data in the surface layer (<2 m) were roughly equivalent, comprising approximately 400 sample points, predominantly dispersed in the LDB. Notably, in the middle (2–15 m) and bottom (>15 m) layers, the numbers of temperature observations were greater than the numbers of salinity and DO measurements, which is attributable to the inclusion of temperature data along the mid-transect. The careful selection of a spatial interpolation method was imperative due to the non-uniform spatial distribution of data and the limited sampling points for the middle and bottom layers. In addition, the average daily wind data in August from 2016 to 2018 used in this article were acquired from the ECMWF ERA-Interim dataset, and the horizontal resolution was 1/8° × 1/8°. The data regarding the daily average current in August from 2016 to 2018 were obtained from the HYCOM global analysis, which was vertically divided into 20 layers (0–100 m) with a horizontal resolution of 1/12° × 1/12°.

2.2. Spatial Interpolation Methods

It is practically impossible and unnecessary to obtain spatiotemporal information on any given continuous phenomenon at every point within a given geographic area. The most practical approach has always been to obtain information about the phenomenon at as many sample points as possible within the given geographic area and estimate the values of the unobserved points from the values of the observed points through spatial interpolation [30].

To determine the best spatial interpolation method for the DO concentration data in the Bohai Sea, we compared five interpolation methods: the RBF-Linear fitting method [33], the inverse distance weighted (IDW) interpolation method [34], the Kriging interpolation method [30], the nearest-neighbor interpolation (NNI) method, and the linear triangular interpolation (LTI) method [35]. The specific calculation process is as follows.

2.2.1. Radial Basis Function Fitting Method

The observed dataset is recorded as D(x_i, y_i, z_i), i = 1, 2, ..., n, where (x_i, y_i) is obtained from the longitude and latitude of point i transformed into a plane rectangular coordinate system (using a Miller map projection). z_i is the depth, and n is the number of observation points. Assuming isotropic distribution of ocean elements, the formula for calculating points $\tilde{D}$(x_j, y_j, z_j) to be interpolated is as follows:

$$\tilde{D}\left(x_j,y_j,z_j\right)={\sum }_{i=1}^n{\beta }_{ij}\phi \left(d_{ij}\right)+{\lambda }_1+{\lambda }_2{x}_i+{\lambda }_3{y}_i+{\lambda }_4{z}_i$$

(1)

The values of β_i, λ₁, λ₂, λ₃, and λ₄ are calculated as follows:

$$\left[\begin{array}{c}\beta \\ \lambda \end{array}\right]={\left[\begin{array}{cc}A& P\\ P^T& 0\end{array}\right]}^{-1}·\left[\begin{array}{c}D\\ 0\end{array}\right]$$

(2)

where

$$\beta =\left[\begin{array}{c}\begin{array}{c}{\beta }_1\\ {\beta }_2\\ ⋮\end{array}\\ {\beta }_n\end{array}\right], \lambda =\left[\begin{array}{c}\begin{array}{c}{\lambda }_1\\ {\lambda }_2\\ {\lambda }_3\end{array}\\ {\lambda }_4\end{array}\right], P=\left[\begin{array}{cc}\begin{array}{c}1\\ \begin{array}{c}1\\ \begin{array}{c}⋮\\ 1\end{array}\end{array}\end{array}& \begin{array}{c}{x}_1\\ x_2\\ \begin{array}{c}⋮\\ x_n\end{array}\end{array}\end{array} \begin{array}{cc}\begin{array}{c}{y}_1\\ y_2\\ \begin{array}{c}⋮\\ y_n\end{array}\end{array}& \begin{array}{c}{z}_1\\ z_2\\ \begin{array}{c}⋮\\ z_n\end{array}\end{array}\end{array}\right], D=\left[\begin{array}{c}\begin{array}{c}D\left(x_1,y_1,z_1\right)\\ D\left(x_2,y_2,z_2\right)\\ ⋮\end{array}\\ D\left(x_n,y_n,z_n\right)\end{array}\right]$$

(3)

$$A=\left[\begin{array}{ccc}\begin{array}{c}\begin{array}{cc}0& \phi \left(d_{12}\right)\end{array}\\ \begin{array}{cc}\phi \left(d_{21}\right)& 0\end{array}\end{array}& \begin{array}{c}\phi \left(d_{13}\right)\\ \phi \left(d_{23}\right)\end{array}& \begin{array}{c}\begin{array}{cc}\dots & \phi \left(d_{1n}\right)\end{array}\\ \begin{array}{cc}\dots & \phi \left(d_{2n}\right)\end{array}\end{array}\\ ⋮& \ddots & ⋮\\ \begin{array}{cc}\phi \left(d_{n1}\right)& \phi \left(d_{n2}\right)\end{array}& \cdots & 0\end{array}\right]$$

(4)

where d_ij is the Euclidean distance between (x_i, y_i, z_i) and (x_j, y_j, z_j). φ(d_ij) is the basis function. In this study, the RBF-Linear fitting method was chosen, that is, $\phi \left(d_{ij}\right)=d_{ij}$.

However, the distribution of ocean elements is anisotropic, meaning that vertical variation is larger than horizontal variation. Through some tests, we increased the weight of the change rate in the vertical direction (z_i′ = z_i/1000) in the calculation process, aiming to keep the change rate of all dimensions as consistent as possible to obtain better interpolation results.

2.2.2. Inverse Distance Weighted Interpolation Method

The IDW interpolation method determines the weight according to the distance between the observation data D(x_i, y_i, z_i) and the interpolation points $\tilde{D}$(x_j, y_j, z_j) and calculates the values of the points to be interpolated. (x_i, y_i) is obtained from the longitude and latitude of point i transformed to the plane cartesian coordinate system (using a Miller map projection). z_i is the depth, and n is the number of observation points. The formula for calculating points $\tilde{D}$(x_j, y_j, z_j) to be interpolated is as follows:

$$\tilde{D}\left(x_j,y_j,z_j\right)={\sum }_{i=1}^n\frac{D\left(x_i,y_i,z_i\right)}{{d_{ij}}^{\alpha }}/{\sum }_{i=1}^n\frac{1}{{d_{ij}}^{\alpha }}$$

(5)

where $d_{ij}=\sqrt{{(x_j-x_i)}^2+{(y_j-y_i)}^2+{(z_j-z_i)}^2}$. α was set to 3.

2.2.3. Other Interpolation Methods

In this study, Kriging interpolation was implemented using a MATLAB toolbox. The specific settings were as follows: the initial guess for the relevant parameters θ (temperature, salinity, and DO) was set to 0.1, and the upper and lower boundaries were set to 10⁻³ and 20, respectively.

In addition, the NNI and LTI interpolation methods were implemented using the ‘scatteredInterpolant’ command in MATLAB.

2.2.4. Tenfold Cross-Validation

To evaluate the interpolation results, 10-fold cross-validation [36] was utilized. The ocean element data were arranged according to latitude and longitude and divided into ten subsets. Nine of these subsets constituted set A, and the remaining subset was set B. The data in set A were used to interpolate values, and the accuracy of estimation was evaluated with the data in set B. The standard deviation (SD) and root mean square deviation (RMSD) were used to evaluate interpolation performance. In addition, the correlation coefficient and spatiotemporal distribution between the interpolation results and the actual observed values were compared in order to determine the most suitable interpolation method.

2.3. Depth of Pycnocline

To estimate the depth of water stratification, we utilized the Brunt–Vaisala Frequency (N), which is determined by the following formula: $N^2=-\frac{g}{\rho }\frac{\mathsf{\partial }\rho }{\mathsf{\partial }z}$, where g is the acceleration due to gravity (−9.8 m/s), ρ is the seawater density, and z is the water depth. In previous studies [16,20], N² values exceeding 10⁻³ s⁻² were used to indicate the presence of strong stratification and a stable water column. Typically, the pycnocline, which is the region of water characterized by a rapid change in density with depth, has a certain thickness. The stratified interface in close proximity to the surface layer is known as the upper pycnocline boundary, while the interface near the bottom layer is referred to as the lower pycnocline boundary.

2.4. Wind Stress and Wind Stress Curl

The calculation formulas for wind stress are as follows [37]:

$$\{\begin{array}{l}{\tau }_x=Rho\times C_d\times Speed\times u\\ {\tau }_y=Rho\times C_d\times Speed\times v\end{array}$$

(6)

where Rho denotes the air density (which is considered to be 1.2 kg/m³), while the value of C_d is a nonlinear coefficient based on Large and Pond [38] used to correct for low wind speed [39]. u and v denote zonal and meridional wind components respectively (unit: m/s), with Speed being computed by taking the square root of the sum of squares of the u and v components. The wind stress curl is calculated by the formula $curl=\mathsf{\partial }{\tau }_y/\mathsf{\partial }{\tau }_x-\mathsf{\partial }{\tau }_x/\mathsf{\partial }{\tau }_y$. The unit of wind stress curl is N/m³.

2.5. Horizontal Transport Rate

The horizontal transport rate of DO is estimated by the formula:

$$F={\displaystyle \iint D{O}_{con}{v}_n}dA$$

(7)

where DO_con denotes the DO concentration, v_n is the velocity component perpendicular to the cross section, and dA is the cross-section area.

3. Results and Discussion

3.1. Interpolation Result Evaluation

In order to streamline the calculation process, this study conducted an evaluation of five interpolation methods utilizing data from August 2017. The results of the cross-validation for temperature, salinity, and DO for each interpolation method are depicted in Figure 3. The cross-validation results for temperature revealed that the SD of the Kriging interpolation method was the lowest on the whole, but its RMSD was the highest. Furthermore, the SD and RMSD values of the other four interpolation methods were very close, indicating that the effect of the four interpolation methods was consistent for interpolating temperature. Regarding salinity, the statistical error findings for the five interpolation methods were closely matched; however, the Kriging interpolation method displayed substantial fluctuation in error across the ten experiments. With respect to DO, the NNI interpolation method demonstrates a high SD and RMSD, while the Kriging interpolation method exhibits an elevated RMSD. In contrast, the statistical errors of the remaining three interpolation methods were comparable. To simplify the analytical process, it is imperative to select a spatial interpolation method that is well suited for all three elements. The SD and RMSD outcomes indicated that, among the five interpolation methods, RBF-Linear, IDW, and LTI yielded similar interpolation effects for temperature, salinity, and dissolved oxygen, with relatively minimal statistical errors.

Figure 4 presents the outcomes of the correlation analysis between the observed values and the interpolation results of each of the three selected methods, serving as an additional means of evaluating their performance. The findings indicated a notably high correlation coefficient between the temperature interpolation results and observation results, followed by DO and, finally, salinity. Specifically, the salinity interpolation results generated using the LTI interpolation method exhibited a weak correlation with the observed data, indicating that the suitability of LTI interpolation method for salinity interpolation may be questionable. Moreover, the disparity between the regression line and the 1:1 line was more prominent for the IDW interpolation method than for the other two methods, and its interpolation results were overestimated in the low-value region and underestimated in the high-value region. According to the above analysis, the RBF-Linear fitting method exhibits more accuracy in generating interpolated results for temperature, salinity, and DO. Therefore, this method was adopted for the three-dimensional interpolation of DO concentration in August from 2016 to 2018.

3.2. Spatial Distribution of DO, Temperature, and Salinity in Bohai Sea in Summer

According to Section 3.1, the RBF-Linear fitting method was found to be the best. In an effort to visually elucidate the disparities between the fitting outcomes and the observed results, the spatial distribution of the mean absolute error (MAE) for both is depicted in Figure 5. The MAE results of the surface, middle, and bottom layers collectively evinced that the high MAE values of temperature, salinity, and DO predominantly manifested in the coastal region, with the smallest MAE observed in the central Bohai Sea. Drawing from antecedent scholarly inquiries [6,15,16], it was established that the region of oxygen deficit within the Bohai Sea primarily concentrates in its central Bohai Sea. In summary, the error in the central Bohai Sea is acceptable, and it is reasonable to use the fitting results to study the oxygen deficit in the central Bohai Sea.

The three-dimensional spatial interpolation outcomes for temperature, salinity, and DO in August from 2016 to 2018 were derived through the utilization of the RBF-Linear fitting method (Figure 6).

Figure 6a–c and Figure 7a–c present a comprehensive overview of the temperature environment of the Bohai Sea in August from 2016 to 2018. The study revealed that the average surface temperature slightly varied from 24 °C to 30 °C, and the temperature of coastal waters was significantly higher than that of the central Bohai Sea. A distinct feature of the central Bohai Sea’s bottom layer was the presence of low-temperature water, which was primarily concentrated on the north and south flanks of the central ridge. At the same time, it was connected to the low-temperature water in the Bohai Strait that was invaded by low-temperature, high-salt water from the Yellow Sea [19], forming the cold-water area in the bottom layer of the Bohai Sea.

Figure 6d–f and Figure 7d–f illustrate the salinity environment of the Bohai Sea in August from 2016 to 2018. In general, the study revealed that the salinity of the central Bohai Sea exceeded 31 PSU. In contrast, in estuarine regions, such as the Yellow River and Haihe River estuaries, the salinity level dropped significantly to around 24 PSU. The spatial distribution of salinity in the bottom layer was essentially identical to that in the surface layer. Notably, the study showed that the salinity in the bottom layer of the central Bohai Sea was greater than 31 PSU, whereas the salinity in the bottom layer near the estuary was relatively low, averaging around 28 PSU.

Figure 6g–i and Figure 7g–i show the spatial distribution of DO concentration levels in the surface and bottom layers of the Bohai Sea in August from 2016 to 2018. The results indicated that the DO concentration in the surface of the Bohai Sea ranged from 6 to 10 mg/L, with areas proximate to estuaries and mariculture farms (Figure 1) exhibiting DO levels exceeding 8 mg/L. The DO concentration in the bottom layer was obviously lower than that in the 100% saturated state, meaning that there was an oxygen deficit in the bottom layer. In particular, the lowest measured DO concentration (3.42 mg/L) was observed in 2017 off the YRE. Across the three years, two areas showing a DO concentration less than 4.5 mg/L were identified: one to the northeast of the YRE and the other to the east of the QHD. This pattern may be attributable to the weak stratification and anticyclone circulation of the central ridge (the lighter-colored part of the central Bohai Sea in Figure 1) weakening the connectivity between the north and south oxygen-deficit regions, eventually leading to the emergence of two core oxygen-deficit regions [19]. At the same time, it is worth noting that the oxygen deficit area in 2018 was more northerly [27]; thus, the mechanism is worthy of further investigation.

To further investigate the relationship among DO, temperature, and salinity in the bottom of the Bohai Sea, we selected a mid-transect (Figure 1b) that traversed the oxygen-deficient region. Since there were only three vertical DO observation sites, the vertical structure of the DO interpolation results was not readily apparent (Figure 8a–c). Nevertheless, the DO concentration still demonstrated a clear stratification pattern, with the surface layer having a higher DO concentration than the bottom layer. The lowest DO values were distributed in the bottom water of the two grooves, and the bottom DO concentration was significantly lower in 2017 than in the other two years, with the lowest DO concentration below 4 mg/L, which was basically consistent with the research results of Zhao et al. [40]. Compared with the DO concentration in 2015 (2.09 mg/L) [6], the DO concentration in this paper was consistently higher than 3 mg/L. However, the study of Chen et al. [41] showed that the DO concentration in the central Bohai Sea from 2019 to 2021 was 2.44-3.7 mg/L, which indicates that the DO deficit in the central Bohai Sea has eased in recent years.

Figure 8d–f illustrates the vertical distribution of temperature in the central Bohai Sea. In the summer, two grooves in the central Bohai Sea exhibited a distinct thermocline, and a transverse thermal front was formed with the surrounding seawater. The salinity demonstrated considerable vertical consistency (Figure 8g–i) except for the low surface water salinity observed in 2018. Furthermore, a lateral salinity front near the Yellow River estuary (approximately 119° E) was observed, but it was distant from the low-oxygen-value area.

Based on the horizontal distribution of DO concentration, temperature, and salinity in the Bohai Sea from 2016 and 2018, the high value of DO concentration was mainly distributed in the high temperature and low salinity water off the estuary. This phenomenon can be attributed to the inflow of nutrients from rivers, which promotes the growth and reproduction of phytoplankton; the resultant photosynthesis by phytoplankton supplements the DO in the water column [19]. In addition, the core oxygen-deficit region in the bottom layer (DO < 4.5 mg/L) was primarily distributed in the south and north cold-water grooves, whereas the location of low DO concentration (DO < 5 mg/L) was not completely consistent with that of the bottom cold-water region. According to the vertical distribution of DO concentration, temperature, and salinity in the Bohai Sea from 2016 to 2018, the primary area of oxygen deficit was surrounded by a vertical thermocline and transverse temperature fronts in the water column. Under the combined influence of the aforementioned factors, vertical exchange and horizontal transport of DO are constrained [17,18,19]. In conclusion, the spatiotemporal heterogeneity of the bottom oxygen deficit in the Bohai Sea is inextricably linked to physical mechanisms such as the variations of stratification and current.

3.3. Effect of Air–Sea Interactions on Bottom DO

The establishment of significant stratification is a prerequisite for seasonal oxygen loss in the Bohai Sea [18], and the degree of DO loss in the bottom layer changes with the change of water column stratification [19]. Figure 9a shows the pycnocline depth and DO concentration in the bottom layer. In the years 2016 and 2017, Sta. A exhibited a shallower lower boundary of the pycnocline, a lesser pycnocline thickness, and a lower concentration of DO in the bottom layer than Sta. B. Conversely, in 2018, Sta. A displayed a deeper lower boundary of the pycnocline, a lesser pycnocline thickness, and a higher DO concentration in the bottom layer. It could be seen that when the lower boundary of the pycnocline was shallower, the DO concentration in the bottom layer was lower.

The stratification and mixing processes are significantly influenced by air–sea interactions that create disturbances in wind fields, ocean currents, and precipitation patterns [42]. The total amounts of precipitation at Sta. A and Sta. B in the 10 days prior to the observation time (Table 1) of DO concentration from 2016 to 2018 are shown in Figure 9b. The total precipitation in 2016 was minimal and therefore had a negligible impact on the depth of the pycnocline. In 2017, both stations experienced a significant increase in precipitation, with Sta. A receiving marginally more precipitation than Sta. B. However, the difference in total precipitation was insufficient to explain the disparity between the two stations in the depth of the pycnocline. In 2018, both stations experienced a marked increase in total precipitation, with Sta. B receiving five times its 2017 total. Meanwhile, Sta. B received substantially more precipitation than Sta. A, and the corresponding pycnocline depth of Sta. B was shallower than that of Sta. A. Therefore, when a heavy rainfall event occurs, the depth of the pycnocline in the water column rises to form a freshwater layer, which means that the stratification of the water column will become more stable [43], ultimately resulting in an increase in the underlying oxygen deficit.

Wind stress is a crucial driver of upper ocean dynamics, influencing the ocean’s interior through the process of Ekman pumping [44,45]. The magnitude and direction of wind stress can significantly impact the vertical structure and circulation of the oceans. Wind stress curl, the curl of the stress vector, is a crucial parameter that determines the direction and strength of the Ekman pumping. When the wind stress curl is greater than zero, it results in downward Ekman pumping, while a negative wind stress curl leads to upward Ekman pumping.

Figure 9c depicts the wind stress and wind stress curl of Sta. A and Sta. B in August from 2016 to 2018. In 2016, the direction of wind stress at Sta. A was southwest, and that at Sta. B was southeast, with the wind stress at Sta. B being greater. In 2017, the direction of wind stress at Sta. A was northwest, and that at Sta. B was northeast, with the wind stress at Sta. A being greater than that at Sta. B. In 2018, the wind stress direction at both Sta. A and Sta. B was southwest, but Sta. A had greater wind stress than Sta. B. Regarding the depth of the pycnocline, the westerly wind stress would cause the pycnocline to become shallower. In addition, the variation in wind stress magnitude in August from 2016 to 2018 at Sta. A and Sta. B was minimal. Therefore, the change in wind stress magnitude in the Bohai Sea has little effect on the pycnocline.

The wind stress curl results show that since both Sta. A and Sta. B had negative wind stress curl values, the Ekman pumping generated in the core oxygen-deficit region was entirely upward. By comparing the depth of the pycnocline and wind stress curl, it was demonstrated that the absolute value of wind stress curl at Sta. A in 2016 and 2017 was greater than that at Sta. B and that the corresponding depth of the pycnocline was also shallower than that at Sta. B. In 2018, although the absolute value of wind stress curl at Sta. A was still greater than that at Sta. B, the corresponding pycnocline depth was greater than that at Sta. B. This can be attributed to the fact that Sta. B’s total precipitation in 2018 was more than 1.5 times that of Sta. A. Based on the aforementioned research, it is evident that an increase in the absolute value of wind stress curl leads to a strengthening of upward Ekman pumping. Consequently, this results in a reduction in the depth of the pycnocline and finally leads to a decline in the concentration of DO in the bottom layer. Simultaneously, this factor contributes to the relatively shallow depth of the pycnocline and the fact that the lowest concentration of dissolved oxygen was observed in the bottom layer in 2017.

In conclusion, the air–sea interaction plays a significant role in determining the concentration of DO in the bottom layer, primarily by influencing the depth of the lower boundary of the pycnocline in the water column. In the oxygen-deficient region of the Bohai Sea, the impact of variations in wind stress magnitude on the depth of the pycnocline’s lower boundary is found to be negligible. However, it is observed that when the wind stress direction is westward or when the wind stress curl reaches a certain threshold, the lower boundary of the pycnocline becomes shallower, resulting in a decrease in the DO concentration in the bottom layer. Generally, the wind stress curl in the oxygen-deficient area off the YRE is larger than that off the QHD, leading to a higher degree of oxygen deficit off the YRE. Nevertheless, heavy rainfall events can disrupt this trend, highlighting that such events are a key factor contributing to the increase in oxygen deficit.

3.4. Effects of Circulation and Horizontal Transport on Bottom DO

Prior qualitative research has indicated that the distribution of DO in the bottom layer of the Bohai Sea is primarily influenced by lateral exchange of DO [19]. Additionally, lateral transport, driven by anticyclonic circulation, promotes an increase in oxygen consumption off the QHD while diminishing oxygen consumption off the YRE [18]. In light of this premise, this study engaged in a discussion regarding the impact of anticyclone circulation on the location of the oxygen-deficit region and proceeded to analyze the effects of current velocity and flow direction on the extent of oxygen deficiency in the two core oxygen-deficit regions. Furthermore, the rates of horizontal transport for these two core oxygen-deficit core regions quantified. Figure 10a–c illustrates the circulation and stream functions in the bottom layer of the Bohai Sea in August from 2016 to 2018. The results indicate summertime anticyclonic circulation in the Bohai Sea, which results in the downward transport of DO [17]. In 2016, an anticyclonic circulation with a central stream function of −0.22 m³/s formed in Liaodong Bay, measuring significantly lower than the surrounding area. At the same time, this means that the center has a higher pressure than the adjacent area, making it difficult for the oxygen-deficient zone in the central Bohai Sea to expand to Liaodong Bay. Meanwhile, the central ridge acts as a barrier, limiting the eastward expansion of the low-oxygen-value area to the east of the oxygen-deficit region. In 2017, an anticyclonic circulation with a very low central stream function developed near the central ridge of the Bohai Sea, resulting in the construction of a standard dual-core structure in the oxygen-deficit region. Increased central pressure propelled the expansion of the oxygen-deficient region outward. In 2018, two anticyclones with minimal stream function appeared in the Bohai Sea, one in the Bohai Bay and the other in the Liaodong Bay. The emergence of the anticyclonic circulation in Bohai Bay and the more northerly anticyclonic circulation in Liaodong Bay led to a substantial northward shift in the oxygen-deficit region. In summary, the variability of anticyclonic circulation in the Bohai Sea significantly influences the location of the oxygen-deficit region.

According to Figure 7, it can be seen that the core oxygen-deficit areas in the central Bohai Sea are mainly located in the sea off QHD and YRE. The difference in oxygen deficit between the two regions is associated with the difference in lateral transport induced by the underlying horizontal circulation [18]. Sta. A and Sta. B are located in two core oxygen-deficit zones, respectively. Sta. A is dominated by a northwest current (Figure 10d–f), with an average current velocity ranging from 0.025 to 0.075 m/s. In 2017, the frequency of current velocities exceeding 0.07 m/s increased compared to the two previous years, highlighting the significant effect of current velocity variation on the DO concentration in the bottom layer off the YRE. Sta. B exhibits greater directional variability in current than Sta. A. In addition to the predominant southwesterly current in 2016, Sta. B was characterized by westerly and easterly currents in 2017 and 2018. Meanwhile, the current velocity of Sta. B remained consistently low, primarily between 0 and 0.05 m/s. Therefore, the change in current direction is a significant determinant of the variation in DO concentration in the bottom layer off the QHD. By comparing the August changes in current velocity, current direction, and DO concentration in the bottom layer at Sta. A and Sta. B, it is possible to conclude that the current mechanism in the bottom layer of the two oxygen-deficient core areas varies. An increase in northwest current velocity exacerbates the decrease in DO concentration in the bottom layer off the YRE, while an increase in easterly current frequency exacerbates the decline of DO concentration in the bottom layer off the QHD.

The variation of current velocity and direction in the bottom layer can affect the DO concentration in the core oxygen-deficit region, which is closely related to the horizontal DO transport. Figure 11 depicts the horizontal DO transport diagram for Sta. A and Sta. B. The current velocity and DO concentration used in this study to estimate the horizontal transport rate refer to the average of the inflow interface and the average of the outflow interface below the lower boundary of the pycnocline. The horizontal DO transport rate through the left cross-section is positive, while the horizontal DO transport rate through the right cross-section is negative. The estimated range is in a circular area with a radius of 0.3° centered on Sta. A and Sta. B. The net horizontal transport rate is the difference between the rate of transport into the circular region and the rate of transport out of it. The results of the net horizontal transport rate estimation are presented in

Table 2. Horizontal transport rate of inflow, horizontal transport rate of outflow and net horizontal transport rate (unit: g/m²/d).

Year	Station	Horizontal Transport Rate of Inflow	Horizontal Transport Rate of Outflow	Net Horizontal Transport Rate
2016	A	4.09 × 103	5.39 × 103	−1.30 × 103
	B	10.76 × 103	11.11 × 103	−0.35 × 103
2017	A	9.18 × 103	11.58 × 103	−2.40 × 103
	B	4.91 × 103	6.99 × 103	−2.08 × 103
2018	A	4.98 × 103	5.55 × 103	−0.57 × 103
	B	5.22 × 103	6.86 × 103	−1.64 × 103

.

Analysis of the estimation results shows a negative net horizontal transport rate in both oxygen-deficit regions over the three years, indicating that the DO supplementation by horizontal transport is less than the loss in the core oxygen-deficit region. Furthermore, the 2017 net horizontal transport rate for Sta. A was substantially lower than the rates for the other two years, confirming that an increase in northwest current velocity exacerbates the loss of dissolved oxygen in the bottom layer at Sta. A. Similarly, the net horizontal transport rates for Sta. B in 2017 and 2018 were notably lower than in 2016, suggesting that an increased frequency of easterly currents will result in greater oxygen deficit in the bottom layer at Sta. B.

Comparing the net horizontal transport rates T Sta. A and Sta. B revealed that in 2016 and 2017, the net horizontal transport rate at Sta. A was substantially lower than that at Sta. B, as was the corresponding DO concentration. In 2018, the net horizontal transport rate at Station A was greater than that at Station B, resulting in a higher DO concentration in the bottom layer than at Station B. In conclusion, horizontal transport is a significant factor influencing the change in DO in the bottom layer and is one of the primary causes of the difference in oxygen deficit between the locations off the YRE and off the QHD.

3.5. The Area and Volume of the Oxygen-Deficit Region

At present, there are water bodies in the Bohai Sea with DO concentrations below 4.5 mg/L, which does not reach the level of hypoxia (DO < 2 mg/L [7]). However, it has been determined that oxygen concentrations below 4.6 mg/L, which correspond to the 90th percentile of the mean lethal concentration distribution, are deemed sufficient to sustain the survival of most species, except for the most sensitive 10%. This oxygen concentration level serves as a precautionary limit that aims to prevent catastrophic mortality events and to effectively conserve marine biodiversity, except for the most sensitive crab species [46]. Therefore, despite the fact that the degree of oxygen deficit in the bottom layer in the central Bohai Sea is lower than that of other low-oxygen regions [47,48,49], it still has a significant impact on the development of coastal fisheries. Numerous research findings have been documented on the extent of oxygen deficit in the Bohai Sea. However, it is noteworthy that only Zhao et al. [19] have undertaken an estimation of the spatial coverage and volumetric magnitude of oxygen deficit in the Bohai Sea in 2014. In order to facilitate a comprehensive analysis of the impact of oxygen deficit in the Bohai Sea, the study conducted an assessment of the spatial extent and volumetric measurements of the oxygen-deficient regions in the Bohai Sea in August from 2016 to 2018. Furthermore, this study examined the potential impact of the depth of the pycnocline and the velocity of the current on the extent and magnitude of the oxygen-deficiency region.

The spatial distribution of areas experiencing oxygen deficit in the Bohai Sea from 2016 to 2018 is depicted in Figure 12a. The variations in the locations of these regions across different years are explained in Section 3.4, attributing these variances primarily to the impact of anticyclonic circulation. The findings regarding the oxygen-deficit area and volume are presented in Figure 12b,c. In 2017, the study also estimated the area and volume of regions with DO levels below 4 mg/L, in addition to levels below 4.5 mg/L. Based on the results shown in Figure 12b, it can be observed that the extent of the oxygen-deficit area, characterized by DO levels below 4.5 mg/L, reached a maximum of 6000 km² in 2017, the highest estimate over the three-year period. In 2016, the oxygen-deficit area exceeded 1778 km², while the minimum estimate in 2018 was approximately 1692 km². Additionally, in 2017, an area of 686 km² was found to have DO concentrations below 4 mg/L, whereas no such area was observed in the remaining two years. Furthermore, upon comparing the areas of the two oxygen-deficit regions in the Bohai Sea, it can be noted that the region off the YRE exhibited a larger area than the region off the QHD. The variations in oxygen-deficit volume (Figure 12c) were generally consistent with the area, except for 2018, when the oxygen-deficit region off the QHD had a higher volume than that off the YRE.

Table 3. The correlation coefficients of oxygen-deficit area/volume with current velocity and mean depth of pycnocline.

	Velocity of Current	Mean Depth of Pycnocline
Area	0.75 (p < 0.08)	−0.78 (p < 0.06)
Volume	−0.01	−0.53 (p < 0.3)

displays the correlation analysis of the depth of the pycnocline and current velocity with the area and volume of the oxygen-deficit region. The findings indicate a positive correlation coefficient of 0.75 between the area of the core oxygen-deficit region and velocity, as well as a negative correlation coefficient of −0.78 between the area and the depth of the pycnocline, with p-values for both correlations approaching 0.05. The limited datasets may be the reason for the high p-values of the above results. Additionally, the study found that the correlations of the volume of the core oxygen-deficit region with the velocity and the depth of the pycnocline were not statistically significant. Overall, the results suggest that the area of the core oxygen-deficit region may increase with higher flow velocity and decrease with greater pycnocline depth. Furthermore, it is worth noting that variables influencing the area and volume of the oxygen-deficit region may also be associated with biochemical parameters, necessitating further investigation.

Overall, there was a significant decrease in both the area and volume of the oxygen-deficit region in the Bohai Sea from 2016 to 2018 compared to the conditions observed in 2014. In 2014, the DO concentration was less than 3 mg/L, covering an area of 756 km² and having a volume of 7820 × 10⁻⁶ m³ [19]. The area of oxygen deficit in the Bohai Sea is strongly influenced by current velocity, while the impact of the average depth of the pycnocline on the area is comparatively less pronounced. However, the influence of current velocity and mean depth of the pycnocline on the volume of oxygen deficit is relatively insignificant. It is noteworthy that the oxygen-deficit area and volume in the Bohai Sea are considerably lower than those in other well-known low-oxygen regions, such as the Changjiang estuary [24] and the Baltic Sea [26]. This observation further supports the notion that the Bohai Sea is currently experiencing a relatively mild stage of oxygen deficit.

4. Conclusions

In order to optimize the computational process, it is essential to utilize a simple and accurate spatial interpolation method to obtain the 3D spatial distribution of various marine elements. This study investigates the spatial interpolation effects of five commonly utilized spatial interpolation methods on temperature, salinity, and DO in the Bohai Sea in August 2017. The results indicate that the RBF-Linear fitting method exhibits favorable attributes in terms of minimal error in temperature, salinity, and DO interpolation. The 3D spatial fields of temperature, salinity, and DO in the Bohai Sea in August from 2016 to 2018 are obtained using the RBF-Linear fitting method, as indicated by the aforementioned research. The spatiotemporal distribution of the three elements shows that the oxygen-deficient region in the Bohai Sea during the summer is primarily distributed in the cold-water zone, which is influenced by the pycnocline and the horizontal temperature front. However, the oxygen deficit in the bottom layer of the Bohai Sea in summer has obvious spatiotemporal heterogeneity. Currently, the analysis of spatial and temporal heterogeneity of oxygen deficit regions primarily emphasizes biochemical and stratification factors, with less focus on physical mechanisms. Therefore, this study further discusses the effects of air–sea interaction, circulation, and horizontal transport rates on oxygen deficit in the bottom layer.

The results of this study highlight the significant influence of wind stress curl and heavy rainfall events on the upward shift of the pycnocline in the Bohai Sea. This upward movement of the pycnocline indicates an intensification of stratification, which indirectly contributes to the expansion of the oxygen deficit in the bottom layer. Additionally, the circulation of anticyclones plays a crucial role in determining the location of oxygen-deficient regions and the variation in the degree of oxygen deficit between the two core areas (off the YRE and off the QHD). The increase in northwest current velocity off the YRE leads to a decrease in the overall horizontal transport rate, resulting in an increase in the oxygen deficit within the bottom layer. Similarly, the heightened frequency of eastward current off the QHD reduces the overall horizontal transport rate, leading to an escalation in the oxygen deficit in that area. Furthermore, the investigation of the oxygen-deficit zone in the Bohai Sea indicates that an increase in current velocity contributes to the expansion of the area of oxygen deficit. Additionally, through a comparative analysis of the degree, area, and volume of oxygen deficit in the Bohai Sea with those of other low-oxygen regions, it has been determined that the Bohai Sea is still in the primary stage of oxygen deficit. In conclusion, these findings contribute to the advancement of our understanding of the physical mechanisms underlying ocean oxygen deficit and may have significant implications for the development of marine fisheries.