**Environmental Influence on the Spatiotemporal Variability of Spawning Grounds in the Western Guangdong Waters, South China Sea**

#### **Yao Lu 1,2, Jing Yu 1,\*, Zhaojin Lin <sup>1</sup> and Pimao Chen <sup>1</sup>**


Received: 22 July 2020; Accepted: 13 August 2020; Published: 15 August 2020

**Abstract:** Spawning grounds occupy an important position in the supplementary population of fishery resources, especially in Western Guangdong waters (WGWs) in the northern South China Sea (SCS), where fishery resources are being depleted. This study investigated the environmental effects on the spatiotemporal variability of spawning grounds in WGWs, on the basis of generalized additive models (GAMs) and central spawning-ground gravity (CoSGG) by using satellite and in situ observations. Results showed that 57.2% of the total variation in fish-egg density in WGWs was explained. On the basis of stepwise GAMs, the most important factor was sea surface salinity (SSS), with a contribution of 32.1%, followed by sea surface temperature (SST), water depth, month, and chlorophyll a concentration (Chl-a), with contributions of 10.7%, 8.8%, 2.6%, and 2.6%, respectively. Offshore distance had slight influence on the model, explaining approximately 0.4% of the variation in fish-egg density. In summary, fish eggs in WGWs were mainly distributed in the area with SSS of 32.0–34.0 Practical Salinity Unit (PSU), SST of 24–27 ◦C, and depth of 0–18 m. CoSGG shifted eastwards by 0.38◦ N and northwards by 0.26◦ E from April to June. The distribution of spawning grounds in the WGW was affected by the Western Guangdong coastal current (WGCC), cyclonic circulation, the SCS warm current (SCSWC), and changes in the habitat environment (such as SST). Fish in WGWs tend to spawn in areas with a high seabed slope and steep terrain (near the Qiongzhou Strait).

**Keywords:** fish eggs; environmental factors; spatial factors; generalized additive model; remote sensing

#### **1. Introduction**

Western Guangdong waters (WGWs) are located in the northern South China Sea (SCS). This is an important place for fish spawning, feeding, breeding, and migration [1]. Spawning grounds are water areas for the mating, spawning, hatching, and breeding of fish, shrimp, and shellfish. It is an important place for the survival and reproduction of aquatic organisms, and it plays an important role in the supplementation of fishery resources [2]. The early life history of fish has three stages: fish egg, larval, and juvenile. The amount of early supplementation and survival rate affects stock density [3]. As an important stage in early-life history, fish eggs are the most vulnerable and sensitive stage in their lives, and small-scale environmental changes may also have a dramatic impact on their resource–replenishment process [4,5]. Previous studies showed that the habitat environment affects the distribution [6] and density [7,8] of fish eggs, the stock density of fish spawning [7,8], and fish structure [9]. However, the influence of seabed topography and marine environment on the spatiotemporal distribution of spawning grounds is still unclear. Spawning grounds are the basis for the replenishment of fishery resources [10]. Exploring the quantitative relationship between the density and distribution of fish eggs and marine environment factors not only helps to understand the formation mechanism of spawning grounds but, also, reflects variations in local fishery resources.

The relationship between fishery resources and marine environment are complex, nonlinear, and nonadditive [11]. In a quantitative analysis of the relationship between fish-egg density and marine environment, the method choice is important. General additive models (GAMs) can better demonstrate the nonlinear relationship between dependent and multiple independent variables [12], and they are widely used in quantitative analyses of the relationship between fishery resources and environmental factors [1,13–15]. On GAMs, the development status of *Clupea harengus* eggs distributed in western Scottish waters was observed to be related to its vertical spatial distribution, and the development of fish eggs near the bottom was relatively slow [16]. In the Baltic Sea, temperature had no significant effect on the abundance of *Platichthys flesus* eggs [17]. The distribution of *Engraulis encrasicolus* and *Sardinella aurita* in the Mediterranean Sea was related to the interaction between seabed depth and sea surface chlorophyll a (Chl-a) concentration. The near-shore continental shelf area with a high sea surface chlorophyll concentration was more suitable for spawning [18]. In China, on the basis of GAMs, the spatial distribution of dominant fish eggs (*Stolephorus commersonnii* and *Cynoglossus joyneri*) in the Haizhou Bay of the Yellow Sea was related to the underlying temperature [19], and the distribution of *Gadus macrocephalus* eggs in the Yellow Sea was closely related to environmental factors such as bottom-water temperature, quality, and salinity [20]. Satellite remote-sensing technology provides all-weather, large-scale, and high-resolution marine-surface information, and it was successfully applied to marine-fishery research [1,13,14,21]. In this study, satellite remote-sensing data were applied in the analysis of the environmental effects of the spatiotemporal distribution of spawning grounds in WGWs. The early supplementary mechanism of fishery resources in WGWs was explored, providing a reference for the protection of fish habitats in the SCS.

#### **2. Materials and Methods**

#### *2.1. Fishery Data*

Fish egg data were obtained from spawning-ground surveys from 2014 to 2015 (April–June). Major species of fish eggs identified in this study were *Trichiurus haumela*, Carangidae, *Nemipteras virgatus*, *Sardinella aurita*, and *Anchoviella commersonii.* The research area was at 110–113◦ E, 19.5–22◦ N (Figure 1). Fish eggs were sampled by macroplankton nets with a hauling speed of 1.5 n mile/h and then preserved in 5% formaldehyde solution. The fish eggs were identified by morphological characteristics, including shape, size, chorion, yolk, oil globule, and pigmentation [9,10]. In this study, fish-egg data were grouped by 0.25◦ × 0.25◦ grid cells. The unit of fish-egg density was ind (10 m)<sup>−</sup>3.

**Figure 1.** Research area and survey stations of spawning grounds in Western Guangdong waters (WGWs; dotted box, area of satellite data).

#### *2.2. Environmental Data*

Satellite data were sea surface temperature (SST), sea surface chlorophyll a concentration (Chl-a), and sea surface salinity (SSS). SST and Chl-a data were derived from MODIS Aqua products of NASA (http://oceancolor.gsfc.nasa.gov), for which temporal resolution was 8 days, and spatial resolution was 4 km. SSS data were obtained from the Global Ocean Physical Reanalysis Product of the Copernicus Marine Environment Management Service (CMEMS, http://marine.copernicus.eu/), for which temporal resolution was one month, and spatial resolution was 1/12◦. The digital elevation model (DEM) of WGWs was derived from elevation data of Google Earth, with an elevation level of 18 and a spatial resolution of 8.85 m.

Satellite remote-sensing SST, Chl-a, and SSS data were derived by removing invalid values and performing monthly averaging and data-fusing by using MATLAB (MathWorks, Natick, MA, USA) software. Satellite remote-sensing SST, Chl-a, and SSS data were processed to monthly images through ArcGIS 10.5 (Esri, Redlands, CA, USA)software (ordinary kriging) [22–24]. The distribution of seabed depth (elevation) and seabed-terrain slope was plotted through ArcGIS 10.5 software by DEM data.

#### *2.3. GAMs Fitting Procedures*

The GAMs is an additive model that was proposed by Hastie [12]. It is a nonparametric method of generalized linear regression [19]. The primary formulation of this model is

$$\mathcal{Y} = a + \sum\_{j=1}^{n} f\_i(\mathbf{x}\_j) + \varepsilon \tag{1}$$

where Y, fish-egg density (ind (10 m)<sup>−</sup>3); xj, explanatory variable (environmental factors for each survey station); α, formulation intercept; ε, residual; and fi(xj), any univariate function of the respective variable with spline smoothing. The formulation of the GAM is

$$\begin{array}{c} \log(\text{Y}+1) = s(\text{Month}) + s(\text{Lon}) + s(\text{Lat}) + s(\text{SST}) + s(\text{Cbl}-a) + s(\text{Slope}) + s(\text{SSS}) + \\ \qquad s(\text{Distance}) + s(\text{Depth}) + \varepsilon \end{array} \tag{2}$$

where Y is the fish-egg density. In order to prevent the response variable from appearing as zero, we made a logarithmic transformation after Y + 1; s(x), spline-smoothing function of covariate x; *Month*, month; *Lat*, latitude; *Lon*, longitude; *SST*, sea surface temperature; *Chl-a*, sea surface chlorophyll a concentration; *Slope*, seabed terrain slope; *SSS*, sea surface salinity; *Distance*, closest distance from the shore; *Depth*, water depth; and E, model error that obeyed the Gaussian distribution. The mgcv package in R v.3.4.4 software (R Core Team, https://www.r-project.org/) was used to build and test the GAMs [25,26], and a forward-stepwise method was employed to select variables with a significant influence on the model.

The Akaike information criterion (AIC) was applied to check the fitness of the model after adding variables to the model [27]. The smaller the AIC value is, the better the model fit. Generalized cross-validation (GCV) was used to assess predictor variables. The smaller GCV is, the greater the generalization ability of the model [28,29]. The significance and nonlinear contribution of the factor to the nonparametric effect were evaluated by F and chi-squared tests, respectively [30–32].

The formula for calculating the AIC value is

$$AIC = \theta + 2df\varphi \tag{3}$$

where θ, deviation; *df*, effective degree of freedom; and ϕ, variance.

#### *2.4. Center of Gravity of Spawning Grounds*

The center of the spawning-ground gravity (CoSGG) of fish-egg density in each month was calculated with reference to the gravity-center analysis method [1], indicating the spatiotemporal variations of spawning grounds in WGWs. The formula for calculating the CoSGG of spawning grounds is [33]

$$X = \sum\_{i=1}^{K} (\mathbf{C}\_i \times \mathbf{X}\_i) / \sum\_{i=1}^{K} \mathbf{C}\_i$$

$$Y = \sum\_{i=1}^{K} (\mathbf{C}\_i \times \mathbf{Y}\_i) / \sum\_{i=1}^{K} \mathbf{C}\_i \tag{4}$$

where X and Y, CoSGG longitude and latitude; Ci, fish-egg density of fishing area i; Xi and Yi, central latitude and longitude positions of fishing area i; and K, total number of fishing areas.

#### **3. Results**

#### *3.1. GAMs Analysis*

The spatiotemporal and environmental factors selected on the basis of AIC and GCV values were month (*Month*), sea surface temperature (*SST*), chlorophyll a concentration (*Chl-a*), sea surface salinity (*SSS*), distance from shore (*Distance*), and waters depth (*Depth*). In this study, the GAM formulation was

$$\log(Y+1) = s(Mmth) + s(SSS) + s(Depth) + s(SST) + s(Chl-a) + s(Distance) \tag{5}$$

The deviance explained by this model was 57.2%, with R2 of 0.531 (Table 1).

**Table 1.** Deviance analysis for the general additive models (GAMs) fitted to the fish-egg density.


In GAMs, the influence of the spatiotemporal and environmental factors on fish-egg density is indicated by the contributions in Table 2. The most important influencing factor from the selected factors was the SSS with a contribution of 32.1%, followed by SST, depth, month, and Chl-a, with contributions of 10.7%, 8.8%, 2.6%, and 2.6%, respectively. The distance had slight influence on the fish-egg density, explaining 0.4%. As indicated by the ANOVA F-ratio test, all selected factors in the model and fish-egg density showed significant correlations (Pr(F) < 0.05). The chi-squared test evaluates the nonlinear contribution of nonparametric effects, and the lowest value (Pr(chi)) was the best. In the GAMs, the factors of the *SSS*, *SST*, and *Depth* were the best.

AIC, Akaike information criterion; GCV, generalized cross-validation; SSS, sea surface salinity; SST, sea surface temperature; and Chl-a, chlorophyll a concentration.

In GAMs, the influence of spatiotemporal and environmental factors on the fish-egg density were indicated by the contributions in Table 2. The most important influencing factor from the selected factors was the SSS, with a contribution of 32.1%, followed by SST, depth, month, and Chl-a, with contributions of 10.7%, 8.8%, 2.6%, and 2.6%, respectively. Distance had a slight influence on the fish-egg density, explaining 0.4%. As indicated by the ANOVA F-ratio test, all the selected factors in the model and fish-egg density showed significant correlations (Pr(F) < 0.05). The chi-squared test is

a type of test to evaluate the nonlinear contribution of nonparametric effects, and the lowest value (Pr(chi)) was the best. In GAMs, the factors of the *SSS*, *SST*, and *Depth* were the best.


**Table 2.** Contributions of the selected variables in GAMs.

\*\*\*, *p* < 0.001; \*\*, *p* < 0.01; and \*, *p* < 0.05. SSS, sea surface salinity; SST, sea surface temperature; Chl-a, chlorophyll a.; d.f., degrees of freedom; Pr(*F*), *p*-value from an ANOVA F-ration test; and Pr(*chi*), a type of score test to evaluate the nonlinear contributions of the nonparametric effects.

The relationships between the fish-egg density and the chosen factors are presented in Table 2. The environmental factors (SSS, SST, and Chl-a) had the most significant influences on the model, with a total contribution of 45.4% (Table 2). Among those factors, SSS had the greatest influence on the fish-egg density, with a contribution of 32.1% (Table 2). Within the range of 24–30.5 Practical Salinity Unit (PSU), an increase in the SSS had a positive effect on the fish-egg density. As the SSS increased, the confidence interval decreased and reliability increased. Within the range of 30.5–32 PSU, the increase in the SSS had a negative effect on the fish-egg density, with its confidence interval decreasing and credibility increasing. Within the range of 32–34.5 PSU, the SSS increased with the increase in the fish-egg density, reaching its maximum of 34.5 PSU. At the same time, the confidence interval decreased and reliability increased (Figure 2b). The SST contribution to the model was 10.7% (Table 2). Within the ranges of 24–25 and 27–29 ◦C, the fish-egg density increased with the increasing SST, and it reached its maximum at 29 ◦C. Within the ranges of 25–27 and 29–32.5 ◦C, the fish-egg density decreased with the increasing SST, reaching its minimum at 32 ◦C. Within the range of 29–32.5 ◦C, the confidence interval increased with the increasing SST, and the reliability was reduced (Figure 2d). Chl-a had a slight influence on the fish-egg density, with a contribution of 2.6% (Table 2). Within the ranges of 0–4 and 10–21 mg m–3, the fish-egg density showed an upward trend, which increased with an increase in the Chl-a, and the fish-egg density reached its maximum at 21 mg m–3. Within the range of 10–21 mg m–3, the confidence interval increased, and the reliability was reduced. Within the ranges of 4–10 mg m–3 and 21–35 mg m–3, the fish-egg density showed a downward trend. The fish-egg density decreased as the Chl-a increased. The fish-egg density had its minimum at 33 mg m–3; near the minimal value, the confidence interval was large, and the reliability was low (Figure 2e).

Spatial factors (depth and distance) contributed 9.2% to the model (Table 2). Depth contributed 8.8% to the model (Table 2). In the range of 0–18 m, the fish-egg density increased with the increase in depth and reached its maximum at 18 m. Within the range of 18–45 m, the fish-egg density decreased with the increase in depth, reaching its minimum at around 50 m. Within the range of 45–80 m, the fish-egg density increased with the increase in depth, and the confidence interval increased and the reliability decreased (Figure 2c). The contribution rate of distance to the model was 0.4% (Table 2). There was a negative linear correlation between the fish-egg density and distance, and the fish-egg density decreased with the increase in distance. After the distance reached 40 km, the confidence interval increased, and the reliability decreased (Figure 2f).

The contribution of the time factor (month) to the model was 2.6%. The fish-egg density gradually decreased with the increasing month value, reaching its maximum in April, and it remained at a high level in April and May. In June, it dropped significantly, reaching its minimum (Table 2 and Figure 2a).

**Figure 2.** A generalized additive models (GAMs) analysis of the effects of the spatiotemporal and environmental factors on the fish-egg density in WGWs: (**a**) month, (**b**) seas surface salinity (SSS), (**c**) depth, (**d**) sea surface temperature (SST), (**e**) chlorophyll concentration of a (Chl-a), and (**f**) distance. Shadow areas, 95% confidence intervals. Rug plots on the x-axis indicate data density.

#### *3.2. Relationship between Fish-egg Distribution and Environmental Factors*

The spatiotemporal distribution of the fish-egg density and SST in WGWs is shown in Figure 3. The SST of the study area was 22–24 ◦C in April, 24–26 ◦C in May, and 26–30 ◦C in June. Areas with a high value of the fish-egg density in April were concentrated in a sea area with the SST of 22–24 ◦C, in May with the SST of 24–25 ◦C, and in June with the SST of 28–30 ◦C (Figure 3a–c). In June, the fish-egg density was generally lower than that in April and May and mainly distributed in high latitudes (20.5–21.5◦ N) and high-SST (> 28 ◦C) waters (Figure 3c).

**Figure 3.** Relationship between the spatiotemporal distributions of the fish-egg density, water temperature, and salinity in WGWs: (**a)** SST in April, (**b**) SST in May, (**c**) SST in June, (**d**) SSS in April, (**e**) SSS in May, and (**f**) SSS in June.

The SSS gradually increased from April to June in WGWs. The high-SSS area tended to move toward the northwestern area (Figure 3d–f). The area with high fish-egg density was concentrated in waters with an SSS of 33.2–33.5 PSU in April, 33.7–33.8 PSU in May, and greater than 33.9 PSU in June.

#### *3.3. Relationship between Fish-egg Distribution and Spatial Factors*

The spatial-factor analysis (distance, depth, and slope) and fish-egg density in WGWs showed that the fish-egg density in near-shore areas was generally higher than that in offshore areas (Figure 4a). The fish-egg density was higher in shallow waters (10–20 m) and lower in relatively deep waters (40–70 m) (Figure 4a). In this study, the fish-egg density reached its maximum near the Qiongzhou Strait (32070.05 ind (10 m)<sup>−</sup>3), where the seabed slope was higher (seabed slopes > 1◦) than that in other areas (density > 10,000 ind (10 m)<sup>−</sup>3) (Figure 4). The fish-egg density in this area was higher than that in other areas (Figure 4b).

**Figure 4.** Relationship between the fish-egg density and spatial factors in WGWs: (**a**) distribution of water depth and (**b**) distribution of seabed slope; r, a score of Pearson correlation coefficient; and P, *p*-value from Pearson correlation coefficient.

#### *3.4. CoSGG Variations of Spawning Grounds in WGWs*

The CoSGG of the spawning ground in WGW moved from the southwest of the study area to the northeast from April to June (Figure 5a). It was located in the area near Leizhou Peninsula (110.87◦ E, 20.68◦ N, green dot in Figure 5a) in April, moved to the eastern waters (111.20◦ E, 20.72◦ N, red dot in Figure 5a) in May, and shifted to the northern waters (111.25◦ E, 20.94◦ N, yellow dot in Figure 5a) in June. The CoSGG of the spawning grounds varied by 0.03◦ N and 0.33◦ E from April to May and 0.22◦ N and 0.04◦ E from May to June.

**Figure 5.** Changes in the central spawning-ground gravity (CoSGG) of the spawning grounds in WGWs: (**a**) moving track and (**b**) longitude and latitude of the CoSGG.

#### **4. Discussion**

#### *4.1. E*ff*ects of Environmental Factors on Fish-Egg Density*

The results of this study highlighted the importance of examining multiple environmental drivers when assessing fish-egg responses to environmental conditions [34,35]. In particular, interactions between geographical and ecological environmental factors obtained from the satellite remote sensing and fish-egg density were found on the basis of GAMs. Among the selected factors, the SSS had the greatest influence on the fish-egg density, with a contribution of 32.1% (Table 2). Fish egg is an early stage of fish-life history. Therefore, the water environment, especially salinity, was one of the major factors affecting the metabolism of fish eggs [36]. Salinity affects embryos developed through changing the osmotic pressure of fish eggs. The research indicated that the hatching rate and increasing salinity showed an approximately normal distribution trend [37]. A higher or lower salinity hinders the material exchange between fertilized eggs and the surrounding medium, resulting in the reduction of the hatching rate and embryo malformation [38]. This study showed that fish eggs in WGWs were mainly distributed in waters with an SSS of 31.5–34.5 PSU, and the most suitable SSS for fish eggs was 33–34.5 PSU (Figure 2b). There were relatively concentrated spawning grounds for *Trichiurus haumela* and *Nemipteras virgatus* in WGWs, where the suitable salinity ranges were 33.0–34.5 [39] and 33.94–34.92 PSU [40], respectively, consistent with the results of this study. In addition, the salinity affected the vertical distribution of fish eggs in the water. In low-salinity waters, fish eggs tended to accumulate, making them unable to get enough oxygen, which was not conducive to the development of the fish eggs. In high-salinity waters, on the other hand, fish eggs could be suspended or floated in water. This could facilitate oxygen absorption from the surrounding water, improving the hatchability of the fish eggs [38,41].

The SSS in the northeastern part of the study area was significantly lower than that in other waters, and the fish-egg density in this area was generally lower than that in other waters (Figure 3d–f) because of the dilution of the Pearl River estuary. From the GAM analysis, the fish-egg density fluctuated with the increase in salinity in the salinity range of 30–34.5 PSU, partly connected with different salinity-suitability levels for various fish-species eggs. The northern SCS is an area with multispecies fishery resources and a complex composition [42]. Suitable salinity (about 34 PSU) provides spawning grounds for fish of different reproductive habits [43]. In addition, the sea-salinity gradient had a certain effect on the spawning grounds. The salinity gradient in Leizhou Bay in China is small, and the salinity has little effect on fish eggs [44]. The salinity gradient in WGWs was larger, and the salinity had a greater impact on the density and spatial distribution of fish eggs.

The contribution rate of the SST to the fish-egg density was 10.7% (Table 2). Fish eggs in WGWs were mainly distributed in waters with SST of 22–32 ◦C, and the most suitable SST for fish egg survival was 24–30 ◦C (Figure 2d). Studies showed that the suitable temperature for the fish eggs of *Trichiurus haumela* in the SCS was 25–28 ◦C [39], consistent with the results of this study. The effects of temperatures at 25 and 29 ◦C on the fish-egg density had two distinct peak areas (Figure 2d), related to the characteristics of multi-fish-species fishery resources in WGWs [43]. The water temperature is one of the key factors affecting fish metabolism [36,45]. The water temperature affected the number, distribution, and population structure of fish eggs by affecting the adult gonadal development and reproductive migration [46]. The water temperature also had a significant impact on the metamorphosis [6] and hatching speed of fish eggs [47]. The fish-egg density sharply decreased in June, which might have been related to changes in the water temperature. The average water temperature in June was higher (greater than 28 ◦C; Figure 3c), which reduced the survival rate of some eggs that did not tolerate high temperatures, leading to a decrease in the fish-egg density [48]. On the other hand, an increase in the water temperature promotes the development of some fish eggs, shortens their hatching time, and accelerates the speed of fish-egg hatching, which also leads to a decrease in the fish-egg density [38]. In the East China Sea, the fish-egg density in the summer (June) was higher than that in the spring [49], which was different from the results of this study. This may be

related to the difference in the water temperatures in this month. In June, the SST (28–32 ◦C) in WGWs was higher than the SST (20.03–27.13 ◦C) in the East China Sea [49]. Different water-temperature levels affected the density and distribution of the fish eggs. The GAM analysis showed that the contribution rate of Chl-a to the fish-egg density was only 2.6% (Table 2). Chlorophyll had little effect on the fish-egg density, because the eggs were in the endotrophic stage and could not prey [6,50].

#### *4.2. E*ff*ects of Spatial Factors on Fish-Egg Density*

The GAM analysis showed that depth ranked the third in the impact on the fish-egg density, with a contribution of 8.8% (Table 2). The research area was located in the sea with a water depth of 8–55 m. The effect of water depth on the fish-egg density showed a fluctuation trend that first increased, decreased, and then increased (Figure 2c). Water depth was one of the major reasons affecting the spatial distribution of spawning grounds [50]. *Trichiurus haumela* in WGWs mainly spawned in waters with a depth of 40–70 m [39]. *Carangidae* fish spawned in waters with depths lower than 60 m in the spring and summer [33]. *Nemipteras virgatus* mainly inhabited the bottom sediment, with a depth of 60–80 m [40]. The eggs of *Anchoviella commersonii* were mainly distributed in areas with depths of less than 20 m [51]. The eggs of *Sardinella aurita* were mainly distributed in areas with a depth of 10 m [52]. *Anchoviella commersonii* and *Sardinella aurita* of pelagic fish and *Nemipteras virgatus* and *Carangidae* of demersal fish inhabited areas with different water depths. Therefore, the effect of water depth on the fish-egg density in the study area showed a fluctuation trend. In WGWs, the fish-egg density in the deep-water areas was significantly lower than that in shallow–water areas (Figure 4a). The relatively harsh lived-in environments of fish living in deeper waters (less food in shallow waters) and other factors might increase the breeding interval of these fish (annual to perennial). Therefore, the fish-egg density in deep waters was lower than that in other areas [53]. The relevant research found that, in the coastal waters of the North Sea, fish eggs tended to gather in shallow waters (<40 m) and form spawning grounds [50,54]. This was similar to the distribution characteristics of spawning grounds in WGWs. This might be due to the abundance of bait organisms in shallow offshore waters, which can provide an ideal spawning ground for fish [53,55]. Fish eggs cannot swim or move independently [49,56]. Therefore, fish eggs floating in the water are susceptible to the effects of ocean currents. The complex seabed terrain reduced the flow velocity and helped fish eggs to gather in this area. This study showed that there was a significant positive correlation between the fish-egg density and seabed slope (*p* < 0.05; Figure 4b). The Qiongzhou Strait (Figures 1 and 4b) had a high seabed-slope value, and the fish-egg density in this area was also higher than that of other areas (Figure 4). This was because the complex seabed topography provides an ideal environment for fish-spawning communities [19,57]. Previous studies showed that the spawning grounds of the Baltic herring (*Clupea harengus membras*) were also distributed in steep areas with steep slopes on the seabed [58]. Therefore, the seabed slope was one of the conditions for the formation of spawning grounds, and the area with a high seabed-slope value tended to attract fish to spawn.

The distance from shore had the least effect on the fish-egg density, with a contribution of 0.4% (Table 2). The distance of the survey stations in this study was 0-60 km, and the effect on the fish-egg density gradually decreased with the increase in distance (Figure 2f). In this study, the fish-egg density in farther areas was significantly lower than that in offshore areas. This might be connected with the reproductive migration of fish to offshore areas in the spring and the formation of a central spawning ground in offshore areas [49].

#### *4.3. Spatiotemporal Distribution of Fish-Egg Density in WGWs*

The center of gravity of the spawning ground in WGWs moved 0.38◦ E to the east and 0.26◦ N to the north from the spring to summer (Figure 5). Fish eggs cannot swim [56], and their distribution was related to physical oceanographic factors such as currents and tides [49]. The position of the spawning ground was affected by changes in the ocean currents. Ocean currents such as the Western Guangdong coastal current (WGCC) and the South China Sea warm current (SCSWC) existed in the WGWs (Figure 6) [59]. There was a coastal flow from the Pearl River estuary along the coast of Western Guangdong to the southwest, lastly crossing the Qiongzhou Strait into the Beibu Gulf (Figure 6) [60]. From May to August, a cyclonic circulation was formed in the area centered at 20–20.5◦ N and 110.75–110◦ E in WGWs (Figure 6b, c). A cold center was formed in this area, with lower temperatures than those of the surroundings (Figure 3c) [61]. In the outer WGCC area, there is a warm current in the South China Sea (SCSWC) that flows fast northeastward all year round (Figure 6) [2,62]. Therefore, fish eggs were affected by the cyclonic circulation (survey time was in late-May) and drifted eastward to gather northeast of Hainan Island (Figure 6) in April and May. In June, the fish eggs drifted northeastward due to the SCSWC with a fast flow. In WGWs, the fish-egg density in each month showed higher in the west and lower in the east, reaching its maximum in the eastern waters of the Leizhou Peninsula (Figure 6). This was related to the WGCC that flew westward all year round. Fish eggs drifting along the WGCC gathered in the northeastern waters of Hainan Island (Figure 6).

**Figure 6.** Spatial distribution of the ocean currents and fish eggs in WGWs: (**a**) April, (**b**) May, and (**c**) June (arrow direction represents the current direction). Black arrow, western Guangdong coastal current (WGCC); blue arrow, cyclonic circulation; and red arrow, South China Sea warm current (SCSWC).

#### **5. Conclusions**

This study analyzed for the first time the environmental effects of the spatiotemporal distribution of the spawning grounds in WGWs on the basis of satellite remote-sensing and survey data. The most important environmental factor affecting the fish-egg density was the SSS, followed by the SST, depth, month, Chl-a, and distance. The spawning grounds in WGWs were mainly distributed in waters with an SSS of 33.0–34.5 PSU, SST of 24–29 ◦C, and depth of 5–25 m. The complex seabed terrain was conducive to the accumulation of fish eggs. The results of this study were helpful in understanding the spatiotemporal distribution of early supplementary populations of fishery populations and their response mechanisms to environmental changes in WGWs.

**Author Contributions:** J.Y. and Y.L. designed the study. Z.L. collected the spawning ground data. Y.L. analyzed the data. P.C. helped the data collection and analysis. J.Y. and Y.L. wrote the article. All authors have read and agreed to the published version of the manuscript.

**Funding:** This study was supported by the following funds: (1) National Key R&D Program of China (2018YFD0900901), (2) Natural Science Foundation of Guangdong Province, China (2018A030313120), (3) Central Public-interest Scientific Institution Basal Research Fund, CAFS, China (2018HY-ZD0104), (4) R & D Projects in Key Areas of Guangdong Province, China (2020B1111030002), and (5) Special Fund for Basic Scientific Research Business of Central Public Research Institutes (PM-zx703-201904-128).

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Article* **Fish Assemblage Structure Comparison between Freshwater and Estuarine Habitats in the Lower Nakdong River, South Korea**

#### **Joo Myun Park 1,\* , Ralf Riedel 2, Hyun Hee Ju <sup>3</sup> and Hee Chan Choi <sup>4</sup>**


Received: 6 June 2020; Accepted: 3 July 2020; Published: 5 July 2020

**Abstract:** Variabilities of biological communities in lower reaches of urban river systems are highly influenced by artificial constructions, alterations of flow regimes and episodic weather events. Impacts of estuary weirs on fish assemblages are particularly distinct because the weirs are disturbed in linking between freshwater and estuarine fish communities, and migration successes for regional fish fauna. This study conducted fish sampling at the lower reaches of the Nakdong River to assess spatio-temporal variations in fish assemblages, and effects of estuary weir on structuring fish assemblage between freshwater and estuary habitats. In total, 20,386 specimens comprising 78 species and 41 families were collected. The numerical dominant fish species were *Tachysurus nitidus* (48.8% in total abundance), *Hemibarbus labeo* (10.7%) and *Chanodichthys erythropterus* (3.6%) in the freshwater region, and *Engraulis japonicus* (10.0%), *Nuchequula nuchalis* (7.7%) and *Clupea pallasii* (5.2%) in the estuarine site. The fish sampled were primarily small species or the juveniles of larger species at the estuary region, while all life stages of fishes were observed at the freshwater habitats. The diversity patterns of fish assemblages varied greatly according to study site and season, with higher trends at estuarine sites during the warm-rainy season. No significant difference in diversity between freshwater and estuarine sites during the cold-dry season were found. Multivariate analyses of fish assemblage showed spatial and seasonal differences of assemblage structures. Higher effects of between-site variability but not within seasonal variability at each site were observed. Variations in assemblage structures were due to different contributions of dominant species in each habitat. Common freshwater species characterized the fish assemblage in the freshwater region, while marine juveniles were significantly associated with the estuarine habitat. The results from the ecological guild analyses showed distinct ecological roles for freshwater and marine species, and overlapping roles for fish sampled at the fishways. The lower reaches of the Nakdong River are an important ecosystem for both freshwater and marine juveniles. Nakdong River estuarine residents and migrant fishes, however, have been negatively affected by the construction of the weir (gravity dam), due to the obstruction to migration from and to freshwater habitats. Conservation and management policies aimed at minimizing anthropogenic influences on estuary ecosystems should focus on evaluating ecological functions of estuary weirs.

**Keywords:** fish assemblage; lower reaches; Nakdong River; estuary weir; marine juveniles

#### **1. Introduction**

Fish assemblages in coastal habitats influenced by rivers consist of a variety of species with diverse life-history strategies and broad functional categories of freshwater, diadromous and estuarine-dependent species. Lower reaches of river systems are characterized by high biological productivity and diversity [1]. Biological communities in these regions are generally influenced by tides, rainfalls, and anthropogenic disturbances [2–4]. In particular, episodic weather events and anthropogenic stressors have significant effects on the structure and functioning of biological communities through declines of biomass, species richness and abundance [5].

Diadromous fish species in temperate Korean rivers include eels (Anguillidae), salmons (Salmonidae) and sweetfishes (Plecoglossidae) [6–8]. Some coastal fishes, including *Coilia nasus* and *Mugil cephalus*, also exhibit occasional upstream movements from coastal waters to lower river reaches [9,10]. Estuary weirs, however, restrict up or downstream dispersal for migratory fishes due to creation of a sharp salinity gradient, controlling spatial distribution in river catchments [11–13]. Fish assemblages in lower river reaches are greatly affected from anthropogenic structures, such as barriers, dams or weirs at both local and catchment scales [12,14,15], contributing to significant discontinuities especially in diadromous fish assemblages [16].

The Nakdong River estuary is an important spawning and nursery ground for many aquatic animals, supporting a productive fishery [17]. In 1987, an estuary weir was constructed to prevent saltwater intrusion into shallow freshwater habitats. The weir has lowered freshwater flows transporting nutrients and sediments to the estuary, threatening the health of estuarine habitats [13,18], such as shifts in population structure of animal communities. Several studies have reported alteration of animal community structures following construction of dykes, weirs, or any other obstruction in estuary environments [19–21]. Kwak and Huh [20] observed that the dominance of fish assemblages in the coastal Nakdong River estuary shifted from demersal to pelagic species after weir construction. The shifts were mostly due to changes in the sediment depositional dynamics caused by the irregular discharge of freshwater, causing changes in circulation patterns within the estuary. Han et al. [16] also documented the loss of species richness following weir construction. Weirs adversely affected diadromous species by blocking their migration routes, but favored nonnative fishes [16,22].

Although many studies have investigated the community structure of estuarine fishes worldwide, and the impacts of estuary weir on fish assemblage structures [15,23–25], such studies are scarce in areas along the southern coast of Korea. Of the few, Hwang et al. [26] have investigated temporal occurrence patterns of fishes in the western Mangyeong Estuary prior to construction of a seawall in 2010, completely enclosing its lower reaches. Park et al. [27] also reported juvenile fish assemblage in the shallow sandy beach of the lower tip of the Nakdong River estuary. Several previous studies have suggested that seasonal variations in fish abundance are primarily the result of species-specific recruitment [28], with environmental factors as strong determinants, especially water temperature and salinity [29–32]. Because lower river reaches play an important role as the nursery habitats for estuarine residents, juveniles of marine species, and temporary habitats for migratory fishes, research evaluating the impacts of weirs within river reaches needs to be conducted for recommendations on restoring connectivity within estuaries where weirs have been constructed and for determining where weirs should be removed.

Anthropogenic effects on freshwater and estuarine fish assemblages are mostly felt at catchment scales. Research, therefore, needs to focus on the impacts at these large scales [33,34]. Thus, this study aimed to examine variations in species composition and abundance at large spatial scales of fishes inhabiting the lower Nakdong River reaches of Korea by evaluating the ecological function of each habitat and the influence of estuary weir on fish assemblage.

#### **2. Materials and Methods**

#### *2.1. Study Area and Sampling*

The study area comprised of the lower Nakdong River, located in the southeastern part of the Korean Peninsula (Figure 1). The Nakdong River weir was constructed 6 km north from the mouth of the estuary (Figure 1). Three and two stations were established to investigate the spatio-temporal patterns of fish assemblages at the freshwater and estuary regions (two study sites), respectively. Stations extended from the southern end of the Nakdong River to 30 km north, into the freshwater site. Additional samplings were also conducted at the fishways (two pool fish ladder type and a gate type) within weir; pool ladder type (length = 24 m, width = 1.8 m, slope = 2.9◦), gate type (length = 50 m, width = 9 m, slope = n.a.). Sampling depths were approximately 5 m at the freshwater site, and between 10 and 15 m at the estuarine site. The mean tidal range at the study area was 1.2 m for spring tides and 0.4 m for neap tides.

**Figure 1.** Sampling stations from freshwater (A1, A2 and A3) and estuary (B1 and B2) sites investigating the effects of weir on fish assemblages at the lower Nakdong River, South Korea.

Fish samples were collected monthly from June 2010 to May 2011 at the freshwater and estuarine stations, as well as at fishways within the weir. Samplings were grouped on annual mean temperature for further analysis based on seasonality. Two seasons (i.e., cold and warm seasons) were determined based on temperatures around 17 ◦C. Temperature data were obtained for Busan province from the Korea Weather Data Open Portal (https://data.kma.go.kr/). Because warm and cold seasons in Korea coincide with rainy and dry seasons, respectively, the seasons were classified as cold-dry (November–March) and warm-rainy (May–October) seasons. Sampling was conducted using a 15 m long, 13 m wide, 3.5 cm mesh bottom trawl with a 1 cm liner covering the codend. Towing speed of ca. 1 knot for 40 min, covering an estimated area of was 16,000 m2, was conducted at each sampling event. A single tow at each station was carried out during the day (between 09:00 and 12:00) simultaneously at the freshwater and estuary regions. Fishway samplings were conducted at the gate and ladder during flood tides. Fish were sampled, the species recorded, and released. Fishway samplings were conducted only during warm-rainy seasons (between May and October) in both 2010 and 2011. In the ladder type fishways, fishes were collected using cone-traps with 20 mm mesh. A set-net (length 10 m, height 4 m, mesh body 50 mm, mesh bag 20 mm) was installed within the gate-type fishway and fish samples were collected after 12 h. All fish samples caught from fishways were checked for species

names, tagged adult specimens using various tags, and then released alive for further fish tracking study. Both surface (between 1 m and 2 m) and bottom (between 10 m and 15 m) water temperatures and salinities were monitored monthly at each sampling location using a portable instrument (Thermo Scientific Orion 3-star). Immediately after capture, fish samples from stations were stored in ice and transported to the laboratory for processing. Taxonomic classification and species name were checked by FishBase [8].

#### *2.2. Ecological Guild and Habitat Type of Fishes*

Five ecological guilds were identified for this study. Guilds were estuarine residents (ER), freshwater species (FW), marine occasional visitors (MV), marine juvenile (MJ), and diadromous (catadromous, anadromous, or amphidromous) migrants (DM), according to native Korean fishes [7,35]. Fish were further categorized into pelagic and benthic.

#### *2.3. Data Analyses*

Full factorial analyses of covariance (ANCOVA) was used to test the effects of site and depth covarying with the seasonal trends of water temperature. A Bonferroni correction for multiple comparison was used to determine post-hoc significance.

The Shannon–Wiener index (H') was used to estimate community-level diversity [36]. A logarithmic transformation (log(*x* + 1)) of fish abundance (number of specimens) was performed to correct for heteroschedasticity and to reduce the weight of overly abundant species in analyses. Three independent two-way ANOVA with an orthogonal design were used to analyze the spatial (two sites) and seasonal (two seasons) effects (independent variables) on fish species richness, abundance and diversity (dependent variables). Prior to ANOVA analysis, homogeneity of variance was tested using Levene's test [37]. Further test onto the dependent variables, the ANOVA tests were also conducted using a split file analysis, with the dependent variable of season paired with site and vice versa. The split file analysis on the ANOVA allowed the data output to be separated by factor to allow visualization of case subsets.

Further inferential and descriptive analyses were performed to assess abundance trends with spatial and temporal patterns. Permutation multivariate analyses of variance (PERMANOVA) on log(abundance + 1) based on Bray–Curtis similarity matrices were conducted [38]. Analysis factors for the PERMANOVA were site (two fixed levels: freshwater and estuary), station (nested within site, two random levels), and season (two fixed levels: cold-dry and warm-rainy seasons). Similarity matrices were used in a PERMANOVA to test for factor effects. In cases in which PERMANOVA detected a significant difference at the 0.05 level, posteriori pairwise PERMANOVA comparisons were used to determine which interaction terms differed significantly among variables within each level of factors. PERMANOVA assigns components of variation (COV) of differing magnitudes to the main factors and interaction between combinations of main factors. The larger COV indicates greater influence of a particular factor or interaction term on the structure of the data [39]. The non-metric multidimensional scaling (nMDS) ordination technique was used to visualize factor effects. To assess statistical significance among factor levels, a canonical analysis of principal coordinates (CAPs) was used [39]. Correlation coefficients between each factor and the canonical axis were used as evidence for species contributions to observed differences. Individual species with both correlations higher than 0.4 and total abundance larger than 1% were plotted on CAP axes 1 and 2 for additional visualization of results.

Statistical software used was Systat (Systat version 18, SPSS Inc., Chicago, IL, USA) and PRIMER v7 with the PERMANOVA+ module [39,40]. A 0.05 level for statistical significance was used in analyses.

#### **3. Results**

#### *3.1. Environmental Variables*

Water temperature and salinity in the study area varied according to seasonal patterns (Figures 2 and 3). Water temperatures at the freshwater site were the lowest in January (1.5 ◦C) and the highest in September (25.2 ◦C). In the estuary, water temperature ranged from 7.3 to 25.7 ◦C at the surface and from 9.1 to 23.8 ◦C at the bottom, both depths showing similar trends with season. No significant differences of water temperature between stations within each season were detected (ANCOVA; *P* > 0.05). Peak water temperatures were observed in August and September, for both freshwater and estuarine habitats, and minima during January and February (Figure 2). Salinity in the estuarine region varied seasonally, ranging from 5.0 to 34.1‰ at the surface and from 25.1 to 34.3‰ at the bottom. Significantly lower salinity at the surface was observed during the rainy season in the estuarine stations (Figure 3; ANCOVA, *P* < 0.05).

**Figure 2.** Seasonal changes in the surface and bottom water temperatures at freshwater (**A**) and estuary (**B**) regions at the lower Nakdong River.

**Figure 3.** Seasonal changes in the surface and bottom salinities at estuary site at the lower Nakdong River.

#### *3.2. Fish Species Composition*

In total, 20,386 individuals belonging to 78 species and 41 families were collected during the study period. Cyprinidae (10 species), Callionymidae (five species), Engraulidae (five species), and Clupeidae (four species) were the most widely represented families (Table 1 and Table S1). Numerically, the five dominant species at the freshwater site were *Tachysurus nitidus*, *Hemibarbus labeo*, *Chanodichthys erythropterus*, *Pseudogobio esocinus* and *Squalidus gracilis*, accounting for 98.0% of the total catch. *Engraulis japonicus*, *Nuchequula nuchalis*, *Clupea pallasii*, *Trachurus japonicus* and *Favonigobius*

*gymnauchen* were the most numerically abundant estuarine fish species, making up 84.7% of the total abundance. Most of the specimens of the dominant species were juveniles of marine fishes, but only *Fa. gymnauchen* was an estuarine resident (Supplementary Material Table S1). In terms of richness, 68 species occurred in the estuarine habitat, which was approximately 3.3–3.5 times more speciouse than freshwater habitats (18 species) or fishways (19 species; Table 1). Only three species co-occurred in both freshwater and estuarine habitats (*Coilia nasus*, *Lateolabrax japonicus* and *Mugil cephalus*), with *Co. nasus* being the most common. In addition, 9 and 12 fishway species co-occurred with fish species in freshwater and estuary habitats, respectively.

**Table 1.** List of fish species, ecological guild (EG), habitat type (HT), and occurrences of fishes in freshwater (Fr), estuarine habitats (Es), and estuary weir fishway (Fw) stations in the lower Nakdong River, South Korea; ER = estuarine residents, FW = freshwater species, MV = marine occasional visitors, MV(J) = MV included juveniles, MJ = marine juvenile, MD = diadromous migrants (DM); B = benthic, P = pelagic.


#### *3.3. Spatial and Seasonal Variation in Species Richness, Abundance and Diversity*

Mean species richness, abundance, and diversity varied by study site and season (Figure 4). Mean species richness at the estuary was higher during the warm-rainy than the cold-dry season, whereas no differences were found between the two seasons at the freshwater site (*F* = 3.724, *P* < 0.05). The mean abundances of fishes were higher during the warm-rainy season at both the freshwater and estuary, with the lowest value during the cold-dry season at the estuary (*F* = 2.182, *P* < 0.05). Diversities also varied between study sites. A higher diversity was found at the estuary than at the freshwater site during the warm-rainy season only (*F* = 0.054, *P* < 0.05).

**Figure 4.** Variations in mean species richness, abundance, and diversity of fish assemblages with study site and season in the lower Nakdong River, South Korea.

The two-way ANOVA confirmed diversity patterns, in that there were significant effects of the site, but not season (Table 2). The two-way interactions were also significant for species richness and diversity, with the exception of abundance (Table 2). Split file ANOVAs showed significant trends of species richness and diversity with site for the warm-rainy season (*P* < 0.05). No significance for any community variable was observed between sites during the cold-dry season (*P* > 0.05). Abundance differences were significant between the freshwater and estuarine site during the cold-dry season only. Within-site seasonal comparisons showed a significant difference for abundance at the estuary habitat and diversity at the freshwater site (*P* < 0.05).

**Table 2.** Results of two-way analysis of variance (ANOVA) on the number of species, abundance, and diversity of fish assemblages in the lower Nakdong River, South Korea; bold face indicates statistical significance at *P* < 0.05.


Mean abundances of the eight dominant species varied seasonally within each habitat (Figure 5). In the freshwater habitat, the mean abundances of *He. labeo* and *Ps. escocinus* were higher in the warm-rainy season, while *Ch. erythropterus* was abundant during cold-dry season (ANOVA, *P* < 0.05). The estuarine fishes showed tendencies of higher abundance during the warm-rainy season only (*P* < 0.05). No seasonal differences of mean abundances were observed for *Ta. nitidus* in freshwater habitats, and for *Nu. nuchalis* in the estuary region (ANOVA, *P* > 0.05).

**Figure 5.** Seasonal variations in mean abundance of 8 common fish species within each freshwater and estuary habitats in the lower Nakdong River, South Korea; asterisks indicate statistical significance at *P* < 0.05 between seasons within each site.

#### *3.4. Fish Assemblage Structure*

PERMANOVA tests revealed fish assemblages were significantly associated with study site and season, but not station (nested within study site), with the COV of site being the highest, indicating the strongest factor determining variation within samples (Table 3). A statistically significant two-way interaction between site and season was also observed (Table 3). Pairwise comparisons of site and season showed evidence of significant differences in fish assemblage structure between freshwater and estuarine site within each season, and also between cold-dry and warm-rainy season with each of study site (PERMANOVA pairwise tests, all *P* < 0.05).

**Table 3.** Mean squares (MS), pseudo-F ratios, significance levels (P), and components of variation (COV) for permutation multivariate analyses of variance (PERMANOVA) tests using Bray–Curtis similarity matrices from abundance of fish assemblages showing differences in site (Si), station (St, nested within Si), season (Se), and interactions terms in the lower Nakdong River, South Korea; bold letters indicate significance at *P* < 0.05.


The non-metric MDS ordination of similarity of mean fish assemblages depicted a clear visual difference between freshwater and estuarine site along with the nMDS ordination horizontally, with the former and the latter site lying right and left side of the plot, respectively (Figure 6). Conversely, the points during both cold-dry and warm-rainy seasons at the estuarine site were interspersed throughout the nMDS plot. At the freshwater site, the points for the cold-dry season lie toward the bottom of the plot, while those for the warm-rainy season were at the upper plot area.

**Figure 6.** Non-metric multidimensional scaling (nMDS) for fish assemblages from the two study sites during cold-dry and warm-rainy seasons in the lower Nakdong River, South Korea.; Fr = Freshwater, Es = Estuary; WR = warm-rainy season, CD = cold-dry season.

Canonical analyses on principal coordinates were performed on significant interactions as a further test on PERMANOVA analyses. The CAP plot for the site–season interaction showed strong evidence for the factor-indicating group separation between sites, and between seasons within each site (Figure 7). Seven fish species were key in separating the estuarine site from the freshwater site, and four species characterized the fish assemblages in freshwater habitats. *Co. nasus* had an intermediate contribution on both sites during the warm-rainy season (Figure 7). Clear seasonal differences in fish assemblages were found at the estuarine site, showing a strong contribution for six species on the warm-rainy season and only a contribution from *Nu. nuchalis* to trends in the cold-dry season. Weak seasonal classifications in fish assemblages were evident at the freshwater site, with weak trends of species contributions on each season (Figure 7).

**Figure 7.** Ordination plots for canonical analysis of principal coordinates of fish assemblage site–season interactions in the lower Nakdong River, South Korea; Fr = Freshwater, Es = Estuary; WR = warm-rainy season, CD = cold-dry season.

#### *3.5. Ecological Guild and Habitat Type of Fish Assemblages*

The analysis of ecological guilds by the number of species within each site across each season showed a high influence of 12 and 13 freshwater species (FW) during the cold-dry and the warm-rainy season, respectively. Collectively, two diadromous migrants (DM) and one estuarine resident (ER) were recorded at the freshwater site during both seasons (Figure 8). In the estuarine site, 13 and 20 marine occasional residents (MV) occurred during the cold-dry and warm-rainy seasons, respectively. No freshwater species were found at this region. All ecological guilds occurred in fishway samples, with dominant freshwater and marine occasional residents (Table 1), indicating that fishways play a role in connecting freshwater and estuary habitats. Marine juveniles (MJ) were only recorded at fishways and the estuary site, with the greatest number recorded in the latter. Freshwater habitats comprised mostly of pelagic species, with the benthic-species number higher at the estuary and fishway regions.

**Figure 8.** The number of species by ecological guild and habitat type in freshwater (Fr), fishway (Fw) and estuary (Es) sites during cold-dry (CD) and warm-rainy (WR) seasons; ecological guilds were estuarine residents (ER), freshwater species (FW), marine occasional visitors (MV), marine juveniles (MJ), and diadromous migrants (DM).

#### **4. Discussion**

A total of 18 and 63 species were collected from freshwater and estuary habitats, respectively, whereas 19 fish species were recorded from fishways. The number of species and diversity was considerably higher at estuary stations. Higher species richness in estuarine regions, compared with freshwater or blackish habitats, has been reported worldwide [41,42]. De Moura et al. [42] indicated that species richness was significantly higher in channels linking estuaries to freshwater ecosystems, due to the regular inflow of saltwater, allowing access for several marine species. Additionally, higher species richness, especially toward estuary mouths, is strongly influenced by marine processes, supporting a greater number of species [43]. Processes such as salinity gradients within the estuary have been shown to strongly influence fish species richness [44]. Such influences, however, may not be general in fish assemblages for all estuaries. Species richness has also been shown to be influenced by regional and local processes affecting colonization, such as processes affecting connectivity between estuaries and the adjacent marine habitats [45].

Of the 63 fish species recorded in the estuarine stations, five were numerically dominant (*En. japonicus*, *Nu. nuchalis*, *Cl. pallasii*, *Tr. japonicus* and *Fa. Gymnauchen*). Estuarine fishes were predominantly juveniles of marine fishes, indicating the importance of the estuary as a nursery [27,46]. The dominance of juvenile fishes in our observations is in general agreement with other studies worldwide (e.g., [47–50]), as well as in Korea (e.g., [26,27,51]). The greater abundance of juveniles

observed in this study indicated the strong dependence of species on the estuary for shelter, survival, and refuge from predators during early stages of their life cycle [48]. As an additional importance of estuaries, most of the marine species that use the estuary as a nursery ground have commercial and recreational value. Among the dominant marine juveniles, *Cl. pallasii*, *Pa. olivaceus*, *Tr. japonicus* and *Tr. lepturus* are the most important fishery resources in Korea [52].

Eighteen fish species were collected in the upper part of estuary weir, of which 14 species were freshwater residents and two species diadromous migrants (*Co. nasus* and *Mu. cephalus*). Among the dominant freshwater species, *Ch. erythropterus* and *He. labeo* are endemic to the Nakdong River [53,54], and *Ta. nitidus* comprises a minor group of freshwater fishes in the Nakdong River [55]. It is worth noting that *Opsariichthys uncirostris*, the common species in freshwater and estuary habitats of the lower Nakdong River [10], was not a major species observed in this study. The low occurrence of *Op. uncirostris* is likely due to extensive dredging between 2009 and 2011 in the lower Nakdong River. The river dredging caused deteriorated water quality and compromised critical habitats, causing the observed low numbers of aquatic organisms [56]. Highly tolerant species to such disturbances, including *Ta. Nitidus*, were abundant in the current study, unlike more sensitive species, such as *Op. uncirostris* [57].

The estuarine weir of the Nakdong River was constructed in 1987, despite many arguments against it, due to the many potential negative impacts on the ecosystem [58,59]. One such impact may be on the changes of fish assemblage structures. The major change in fish assemblage is a decrease of total species number, especially freshwater fishes after the construction of estuary weir [60]. Changes in fish functional categories were also observed from demersal to pelagic in estuary fish assemblages [20]. In addition, migratory fishes also have shown dramatical reduction of their upstream migration route and relative abundance due to blockage of their passage by the estuary weir [60]. Thus, the role of fishways in determining linkages between freshwater and estuarine habitats may be particularly significant for shaping fish communities, especially communities of diadromous migrants.

The results of this study show similar proportions of fish species within freshwater, marine, and stenohaline (estuarine residents and diadromous migrants) guilds in the fishway samples. Fishways were constructed for the provision of passages for animals migrating from estuary to freshwater habitats. Relatively few diadromous species, however, were found in this study, despite the fishways, even with many diadromous species present in the lower reaches of the Nakdong River [10,61]. Because of the obstruction of the estuary weir in fishes' migration into the upper river, several studies have reported reduced migratory fishes in the Nakdong River estuary [10]. Yang et al. [61] reported fishways in the weir of Nakdong River estuary to have limited use, based on the relatively low species richness of anadromous fishes upriver from the structure. In addition, reduced freshwater flow caused by low precipitation can also exacerbate negative impacts on migratory fishes, due to fishways from weir construction becoming of even less efficiency in the provision of adequate passage [62,63]. Effective alternatives for providing fish passage, such as new fish passage designs, implementing recovery of affected habitats, and implementing adaptive management practices, are urgent and necessary for areas in Korea where fishways exist or are planned.

Overall diversity patterns of fish assemblages did not show significant seasonal patterns. A higher fish abundance, however, in the estuarine site was evident during the warm-rainy season. Seasonal variations in fish abundance are mainly due to the presence of species utilizing the Nakdong River estuary during the warm-rainy season for reproduction. For example, the increase in the abundance of *Cl. pallasii*, *En. Japonicus*, and *Tr. japonicus* occurred mainly during the warm-rainy season for *Cl. pallasii* following their spawning periods [64], and during the spring-early summer for *En. japonicus* and *Tr. japonicus* [65,66]. Those same species are also abundant at adjacent near-coastal waters of southeastern Korea during the spring and summer [56,67]. One may thus conclude that the abundances of those species during the warm-rainy season are likely due to the nursery grounds provided by the estuary, whereas the consistent occurrence of *Nu. nucalis* during all seasons suggests that estuaries are the main habitat for their entire life cycle [20].

#### **5. Conclusions**

In conclusion, the lower reaches of the Nakdong River are an important ecosystem for a diverse array of fish species. Of the 78 species sampled in this study, 57 and 14 originated from marine and freshwater areas, respectively, even with limited access to their habitats by the estuary weir. Most of the species sampled in estuaries were exclusively represented by juveniles, highlighting the importance of this ecosystem as a nursery ground. As in many estuaries across the world, many of the species observed in this study were occasional residents. The effect of obstructions, such as the estuary weir, on their success was reflected by the reduction of species richness in areas upriver of the weir than in the previous observation [60]. Studies such as this, investigating the dynamics of species assemblage in lowland river systems impacted by physical obstructions, are critical for the conservation efforts of habitats supporting ecologically and economically important species.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2077-1312/8/7/496/s1, Table S1: List of species including Families, Scientific names, Common names and their abundances (N) in freshwater and estuary sites, and total abundance as well as percentages (%N).

**Author Contributions:** All authors have read and agree to the published version of the manuscript. J.M.P.; methodology, validation, formal analysis and writing—original draft preparation, R.R.; writing—review & editing; H.H.J.; conceptualization, H.C.C.; investigation and data curation.

**Funding:** This research was funded by the Korea Institute of Ocean Science & Technology [grant numbers PE99813].

**Acknowledgments:** We are grateful to Ki Mun Nam and Dong Jin Lee for assistance with samplings and data analyses. Field surveys were conducted in accordance with the approval of "Research & Training Fishery" in the Ministry of Ocean and Fisheries, Korea.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

#### *Article*
