Effects of Climate Events on Abundance and Distribution of Major Commercial Fishes in the Beibu Gulf, South China Sea

Highlights

What are the main findings?

• The abundance of most of fish stocks in the Beibu Gulf continued to decline over the last 15 years.

What is the implication of the main finding?

• Environmental variations caused by climate events can episodically enhance the abundance of certain fish stocks.
• Warming may be the reason of northward shifts in distribution of most fishes.

Abstract

Improving prediction of ecological responses to climate variability requires understanding how local fish population dynamics are impacted by climate events. The present study was conducted in the Beibu Gulf of the northwestern South China Sea where the fisheries are characterized by high ecological and commercial value. We evaluated the relationship between major commercial fish population dynamics (abundance and distribution) and climate periods, using survey data from 2006–2020. The analysis using random forest and GAM models show that climate events are not the best predictors for the variations of fish abundance, because abundance of most fish stocks decreases significantly with the year, and the increasing fishing pressure over time can better explain the overall downward trend in fishery stocks. However, environmental variables that correlate significantly with interannual variation in ONI may impact fish abundance in short terms. Our research suggests that climate events leading to higher surface seawater salinity in winter favors pelagic fishes by improving habitat availability, and higher near-surface chlorophyll-α concentration during La Niña events provides better food condition for overwintering fish. In addition, there is no clear evidence that climatic events have a significant impact on gravity center of fish distribution, whereas climate change has caused most fishes to move to cooler coastal waters in the north.

1. Introduction

Investigating the abundance and distribution of marine fishes provides vital information to test and improve the theory of ecosystem-based fisheries management. The scale and complexity of the critical information related to fish stock dynamics are important determinants of the stability of fisheries ecosystem [1]. Fish abundance and distribution has been shown to be fundamental to understanding sustainable development of fishery resources [2], likely due to the fact that appropriate population status provides higher ecological capacity for stocks, i.e., increase fish population’s ecological resilience [3]. Changes in the abundance and distribution of most major commercial fish species have often been considered as consolidated results of environmental effect and fishing pressure [4]. Past research has emphasized that information on relationships between fish population and overfishing is essential for explaining the processes driving patterns of fish abundance and distribution [5,6]. However, influences of climate variations have proven to have substantial effects on the increase and/or decrease in the abundance and distribution of fish stocks, and the success of fish stock assessments in the future depends largely on the ability to predict the dynamic impacts of climate change on marine ecosystems [7].

The most obvious driver of interannual climate variability is generally recognized as the El Niño–Southern Oscillation (ENSO) [8]. This cycle of alternating warm (El Niño) and cool (La Niña) events is the dominant year-to-year climate signal on the earth [9]. These climate events exert a powerful influence on the atmospheric and marine environment in the tropical Pacific Ocean. This influence may be extended to the adjacent South China Sea through the atmospheric [10,11] and oceanic teleconnections (i.e., seawater transfer) [12]. ENSO is probably more important than processes of climate change or variability that operate more slowly, which can dramatically alter global oceanic conditions, climate, and weather patterns on short notice [9]. Although there is still a lack of consensus on how ENSO amplitude will evolve in the future, recent studies suggest that ENSO-related extreme climate events may occur more frequently with continued global warming [13,14,15]. Projections by Cai et al. [16] suggest that, with increased frequency of extreme El Niño and La Niña events, synchronizing with faster warming in the eastern equatorial Pacific, the frequency of extreme El Niño events will almost double in the future.

Extreme weather events such as those associated with ENSO could significantly affect marine ecosystems, radically change population numbers, or even destroy them altogether. A prime example is the pulsed heat stress events associated with El Niño events reinforcing the impact of global warming on coral reefs, such as coral bleaching on a global scale [17]. In addition to affecting marine ecosystems through heat transport, ENSO events are the major contributor to interannual variability of tropical cyclone in the Pacific Ocean [18]. The vulnerability across the Pacific coast is primarily affected by ENSO [19]. Overall, climate forcing of ENSO will affect physical processes [20,21], primary production [22,23], and prey–predator relations [24] in local marine ecosystems, and thereby ultimately impact habitats [25], migration patterns [26], population recruitments [27], and food webs [28], thus their effects may experience time lags before being felt via, for instance, changes in fishing efforts. Past investigations have shown that fisheries in the Pacific Ocean were strongly influenced by ENSO events, with average landings decreasing ~0.8 million tons in El Niño and increasing ~1.1 million tons in La Niña years since 1950 [29]. As the margin of the Pacific Ocean, the Yellow Sea and East China Sea marine ecosystem structure has been shown to be affected not only by fishing but also that their decadal variations correspond well with contemporaneous climatic regime shifts in the Pacific [30]. Furthermore, the study by Yin et al. [31] showed that there were obvious variations in the species composition and biomass in the Haizhou Bay ecosystem during the El Niño event, indicating that even in heavily anthropogenic activities impacted waters, the effect of climate change is still significant. Therefore, promoting biological and ecological studies, and improving understanding of the relationship between offshore fishery resources and climate events would contribute to sustainability of offshore fisheries and their adaptive capability to confront climate events in future fishery management.

The Beibu Gulf, in northwest South China Sea, is characterized by great diversity of fishes and abundant species of commercial and ecological importance; it is one of the four most prominent fishing grounds in Chinese offshore [32]. There are ~960 species of fishes, belonging to 475 genera and 162 families, and ~80% of them are demersal fishes and ~20% are pelagic fishes [33]. However, a large number of studies have demonstrated that the pelagic stocks are in a state of overfishing and demersal stocks generally in a state of depletion [34]. While a few fish stocks have increased due to niche replacement and/or reduced competition, abundances of those species with large sizes and long-life span have declined to a low level [35,36]. In addition, as a marginal sea in the west Pacific, the South China Sea is inevitably affected by ENSO events that originate from the equatorial Pacific. Numerous studies have indicated that climate events significantly affect the annual ocean-atmosphere cycle in the South China Sea, including that of volume transport [12,37], heat wave [38], rainfall [39,40], monsoon [41,42], and surface seawater temperature [43,44]. Recent research has shown that small pelagic fish blooms in Beibu Gulf occurred in the summers of La Niña years, presumably because of higher densities, smaller body sizes, and later spawning times following La Niña events [45]. Even though the strength of impacts of climate events on both fish stocks and their associated fisheries depends on the type of events, Peruvian experience in the management of commercial fisheries has demonstrated that adaptive fisheries management is well suited to cope with ENSO impacts [29]. Despite that Chinese government’s fishery management has made some progress in the South China Sea over the past decades [46], in the context of complex future trends of climate change, there is an urgent need to establish ecosystem-based fisheries management that takes into account climate variability.

Based on above description of stock status of the Beibu Gulf, the main goal of this paper is to evaluate effects of climate events on the major commercial fishes. For this purpose, we intent to answer the following questions: (i) Which is the key factor that relates fish abundance in the Beibu Gulf, climate event or fishing pressure? (ii) What is the correlation between fish populations in the Beibu Gulf and climate events or environmental variables? (iii) Do climate events affect distribution of fishes in the Beibu Gulf? These assessments of effects of climate events on major commercial fishes would improve our understanding of how large-scale climate variability influences fishery stocks, advancing our ability to further understand and manage future fishery resources.

2. Data and Method

2.1. Study Site

This study concerns the Beibu Gulf in the northwestern South China Sea. The gulf spans a distance of 300 km from the coast of Beihai City in the north (21°15′ N, 109°30′ E) to the western end of the Hainan Island in the south (17°30′ N, 106°45′ E; Figure 1). It is a semi-close gulf with tropical marine environment but with components of tropical and subtropical fish species. Benefitting from its unique geographic location and climatic conditions, the Beibu Gulf serves as spawning and feeding grounds for many important commercial fishes (Chen et al., 2009; Gao et al., 2017).

2.2. Sampling Protocol and Equipment

Data of fish sampling came from otter trawl surveys at 52 stations (from 2006 to 2020, during which period a total of 34 survey cruises were carried out (

Table 1. A list of sampling date and climate condition.

Cruise No.	Sampling Date (Month-Year)	Seasons	Climate Events	Sampling Station Numbers
1	January 2006	Winter	La Niña	51
2	April 2006	Spring	Normal	52
3	July 2006	Summer	Normal	52
4	October 2006	Autumn	El Niño	51
5	January 2007	Winter	El Niño	52
6	April 2007	Spring	Normal	52
7	July 2007	Summer	La Niña	52
8	October 2007	Autumn	La Niña	52
9	January 2008	Winter	La Niña	51
10	July 2008	Summer	Normal	52
11	January 2009	Winter	La Niña	52
12	July 2009	Summer	El Niño	52
13	January 2010	Winter	El Niño	52
14	July 2010	Summer	La Niña	52
15	January 2011	Winter	La Niña	52
16	July 2011	Summer	La Niña	52
17	January 2012	Winter	La Niña	52
18	July 2012	Summer	Normal	52
19	January 2013	Winter	Normal	52
20	July 2013	Summer	Normal	52
21	July 2014	Summer	Normal	52
22	October 2014	Autumn	El Niño	38
23	January 2015	Winter	El Niño	52
24	April 2015	Spring	El Niño	38
25	July 2016	Summer	La Niña	38
26	November 2016	Autumn	La Niña	38
27	January 2017	Winter	Normal	38
28	April 2017	Spring	Normal	38
29	April 2018	Spring	La Niña	38
30	September 2018	Autumn	El Niño	37
31	April 2019	Spring	El Niño	38
32	September 2019	Autumn	Normal	33
33	April 2020	Spring	Normal	38
34	September 2020	Autumn	La Niña	35

). Although most survey cruises did not cover all the 52 stations for various reasons, the sampling was generally coherent in methodology. The surveys were undertaken using a commercial fishing vessel with main engine power of 600 HP. In every survey cruise, each station was trawled once for 1-h duration with a towing speed of 2.5–3.5 knots. The headrope length the otter trawl net was 37.7 m and cod-end mesh 40 mm. Every trawl sample was sorted onboard and identified to species level, catches of each species being weighed, counted, and recorded, and length and weight frequency data were also collected for major commercial species. Based on the combination of frequency and market value, we selected 15 major commercial fishes with the highest frequencies of occurrence for the present analysis (

Table 2. The 15 major commercial fishes in the Beibu Gulf.

Common Name	Family	Scientific Name	Habitat	Price Category
Pacific rudderfish	Centrolophidae	Psenopsis anomala	Benthopelagic	Very high
Big-head pennah croaker	Sciaenidae	Pennahia macrocephalus	Demersal	Medium
Largehead hairtail	Trichiuridae	Trichiurus lepturus	Benthopelagic	High
Red bigeye	Priacanthidae	Priacanthus macracanthus	Reef-associated	High
Threadfin porgy	Sparidae	Evynnis cardinalis	Reef-associated	Very high
Greater lizardfish	Synodontidae	Saurida tumbil	Reef-associated	Very high
Brushtooth lizardfish	Synodontidae	Saurida undosquamis	Reef-associated	Very high
Glowbelly	Acropomatidae	Acropoma japonicum	Demersal	Unknown
Sulphur goatfish	Mullidae	Upeneus sulphureus	Demersal	High
Japanese goatfish	Mullidae	Upeneus japonicus	Demersal	High
Golden threadfin bream	Nemipteridae	Nemipterus virgatus	Demersal	Very high
Yellowbelly threadfin bream	Nemipteridae	Nemipterus bathybius	Demersal	High
Japanese scad	Carangidae	Decapterus maruadsi	Pelagic	Very high
Japanese jack mackerel	Carangidae	Trachurus japonicus	Pelagic	High
Indian mackerel	Scombridae	Rastrelliger kanagurta	Pelagic	Very high

), include 11 fish family (Centrolophidae, Sciaenidae, Trichiuridae, Priacanthidae, Sparidae, Synodontidae, Acropomatidae, Mullidae, Nemipteridae, Carangidae, and Scombridae.

2.3. Data Sources

We selected 4 environment factors that are commonly recognized to have significant impact on fish stocks, including surface seawater temperatures (SST), near-surface chlorophyll-a (Chl-a) concentration, mixed-layer depth (MLD), and surface seawater salinity (SSS). The SST data with resolution of 1/4° were obtained from the OI SST V2 High Resolution Dataset provided by NOAA’s Physical Science Laboratory, Boulder, Colorado, USA (https://psl.noaa.gov, accessed on 1 October 2022). Data of Chl-a, MLD, and SSS were obtained from the Copernicus Marine Service Information (https://www.copernicus.eu/en, accessed on 1 October 2022) of the EU. The dataset of SSS and MLD are products of the GLOBAL_MULTIYEAR_PHY_001_030, which are global ocean eddy-resolving (1/12° horizontal resolution) reanalysis of the Copernicus Marine Environment Monitoring Service covering varied altimetry and since 1993, and the dataset of Chl-a is the product of the GLOBAL_REANALYSIS_BIO_001_029, which is produced at the Mercator-Ocean for the period of 1993–2019 at 1/4° horizontal resolution.

To characterize ENSO events, we also shaped data of the Oceanic Niño Index (ONI) from products of the NOAA’s Climate Prediction Center (CPC; https://origin.cpc.ncep.noaa.gov/, accessed on 1 October 2022). The ONI is defined as 3-month running mean SST departures in the Niño 3.4 region (5° N–5° S, 120–170° W), and is a principal measure for monitoring, assessing, and predicting ENSO. Climate events are defined as 5 consecutive overlapping 3-month periods at or above +0.5 °C anomaly for warm (El Niño) events and at or below the-0.5 °C anomaly for cool (La Niña) events. The threshold is further broken down into Weak (with a 0.5–0.9 °C SST anomaly), Moderate (1.0–1.4 °C), Strong (1.5–1.9 °C), and Very Strong (≥2.0 °C) events. The CPC considers these anomalies must also be forecasted to persist for three consecutive months.

2.4. Data Processing

Stock densities of all the major commercial fishes were expressed as catch per unit area (CPUA), and CPUA is calculated as follows:

$$CPUA=\frac{C}{v·t·L·X}$$

where, CPUA is the fish abundance (kg/km²); C is catching biomass (kg); v is towing speed (km/h); t is trawling duration (h); L is the headrope length (km) of the otter trawl net; and X is the fraction of headrope length of the trawl net and 0.66 is adopted for this study [47].

Since the fish abundance data in the Beibu Gulf calculated in this study cannot fully cover all seasons in the survey period, this study will use the R package of “imputeTS” to impute missing values. Furthermore, given that fish abundance varies linearly with the year and has seasonal differences, use the R package of “fields” to eliminate annual and seasonal trends by the log transformation of fish abundance. The residuals were then used for correlations with environmental variables. For the environmental variables, the interannual trends and seasonal cycles were also removed. All the data processing was performed with R.

2.5. Data Analysis

To investigate the responses of fish populations to climate events, we used linear models to identify relationships between the ONI and fish abundances and gravity centers of their ranges. Spearman correlation was used to test significance of these relationships and p < 0.05 was considered significant. Furthermore, one-way analysis of variance (ANOVA) was employed to compare the fish abundance and distribution between climate periods and between seasons. SPSS 22.0 (SPSS Inc., Armonk, NY, USA) was used to perform these statistical analyses.

We used random forests (RF) algorithm as included in the package “randomForest” [48] of the R software, to identify the environment variables that best predict fish stock density in different climate periods. Like boosting and bagging trees, RF is considered an “ensemble learning” method [49], a collection or ensemble of classification trees each capable of producing a response when presented with a set of predictor values. We used 500 trees to build the RF because increasing this default number did not substantially change the results of variable importance or explained variation [48]. Variable importance measures were derived based on the difference in prediction accuracy before and after permuting each predictor variable.

We then used generalized additive models (GAMs; additive and continuous models) and the “mgcv” package for R was performed to identify the relationships between fish abundance and environmental variables. GAMs comprise a collection of non-parametric and semi-parametric regression techniques for exploring relationships between responsive and explanatory variables, whose advantage is the assumption of additive and stationary relationships between the responsive and explanatory variables. Models containing variable combinations with variance inflation factor > 10 were excluded, to eliminate potential problems with collinearity. Stock density data were modeled using a quasi-poisson distribution, implemented via a call to the “gam()” function (mgcv). The effective degrees of freedom (def; representing the level of nonlinearity) were limited to 4 for avoiding overfitting.

Spatial and temporal distribution of fishes was evaluated by determining the monthly longitudinal and latitudinal gravity centers. The gravity center of fish range was calculated by the following equation [50]:

$$LA{T}_j=\frac{\sum \left(LA{T}_{i, j}\times D_{i, j}\right)}{\sum D_{i,j}}, LO{N}_j=\frac{\sum \left(LO{N}_{i, j}\times D_{i, j}\right)}{\sum D_{i,j}}$$

where LAT_j and LON_j, respectively, denote the latitude and longitude of the gravity center of fish range in a specific cruise j; D_i,j denotes the abundance of a fish species in the ith station in jth cruise; and LAT_i,j and LON_i,j, respectively, denote the latitude and longitude of the ith station in jth cruise.

3. Results

3.1. Environmental Effects on Fish Abundance

In the RF analysis, ranking of the most important explanatory variables driving the log-transformed stock density of the 15 commercial fishes was obtained. Percentage increases in the MSE (mean squared error) of variables were used to estimate the importance of these predictors, and higher MSE% values imply that the predictors are more important. According to the importance ranking of the predictive variables, the significance of predictor indicated that Year, SSS, SST, Chl-a, and MLD (all p < 0.05) were the important factors for the prediction of those fishes’ stock density (Figure 2), whereas Season and Type of Climate Periods were less important.

The final GAM models were set with the results of RF analysis and collinearity analysis as follows:

log₁₀(CPUA + 1) = a + s(Year) + s(Chl-a) + s(SSS) + s(SST) + s(MLD) + ε

where CPUA (abundance) was transformed by log₁₀(A + 1) to downplay outlying values and better represent relationships with environmental variables; a is the intercept of the model; s () is the spline smoothing function of the variables (Year, Chl-a, SSS, SST, and MLD) (

Table 3. Summary of GAM results for major commercial fishes in the Beibu Gulf.

Species	R2	Deviance Explained (%)	GCV	Approximate Significance of Smooth Terms
				Year	Chl-a	MLD	SSS	SST
P. anomala	0.303	29.7	1.3052	***	***	**	***	***
P. macrocephalus	0.259	32.9	1.8712	***	***	ns	***	***
T. lepturus	0.240	21.3	1.1757	***	***	***	**	ns
P. macracanthus	0.259	32.2	0.7861	*	***	***	***	***
E. cardinalis	0.152	14.8	1.4088	***	***	***	***	***
S. tumbil	0.174	18.6	1.1200	*	***	***	***	***
S. undosquamis	0.228	24.2	1.0546	***	***	***	***	***
A. japonicum	0.241	25.3	2.4356	*	**	***	***	***
U. sulphureus	0.210	23.2	1.0214	***	ns	***	***	***
U. japonicus	0.049	13.2	0.6859	ns	**	***	***	***
N. virgatus	0.287	32.3	0.8922	***	**	**	***	***
N. bathybius	0.139	27.9	0.6002	***	**	ns	***	**
D. maruadsi	0.235	26.2	1.4142	***	***	***	***	***
T. japonicus	0.268	26.6	1.7240	***	***	***	***	***
R. kanagurta	0.100	16.3	0.8901	**	***	*	***	ns

Note: GCV: Gross Calorific Value; ns (p > 0.05), * (p < 0.05), ** (p < 0.01) and *** (p < 0.001).

); ε is the random error term of normal distribution and with zero mean and finite variance.

The GAM plots (Figure 3) show the best fitting smoothers for effects of temporal and environmental variables on fish abundance. These plots indicate 95% confidence limits and narrow confidence limits indicate high relevance and vice versa. Abundances of the 15 major commercial fishes are significantly correlated with the Year, and abundances of 11 out of the 15 species have decreased in recent years. The GAM models also indicate that abundances of most fish species do not increase significantly with Chl-a, but are positively correlated with SSS, and the peak of abundances of most species appeared within MLD ranges of 30–50 m. In addition, abundances of these species have varied responses to SST, for example, abundances of P. anomala, T. lepturus, and A. japonicum apparently increase with SST, while S. tumbil, S. undosquamis, D. maruadsi, and T. japonicus have bimodal relationships with SST.

3.2. Changes in the Fish Abundance and Environmental Variables over Time Series

The SSS and SST slightly decreased and then increased over the time series (Figure 4). In addition, the SST and Chl-a with seasonal cycles removed were both significantly correlated with the ONI during 2006–2020 (r = 0.393 and -0.191, respectively, both p < 0.05), but SSS was significantly positively correlated with the absolute value of ONI (r = 0.108, p < 0.05).

In addition, because abundances of most fishes show downward trends with year presumably due to increasing fishing pressure, we detrended log-transformed abundances of fish for further analysis (Figure 5). Although there was no evidence that abundances of all fishes varied between climatic events (p > 0.05), abundances of most fishes increased significantly after any climatic event, exhibiting seasonal peaks and troughs. After moderate or strong (climate) events, the abundance of some demersal (e.g., P. anomala and E. cardinalis) and pelagic fish (e.g., D. maruadsi and T. japonicus) generally increased significantly compared, while some fish species (e.g., S. tumbil, and S. undosquamis) experienced higher seasonal peaks in abundance during ENSO periods.

3.3. Climate Effects on Gravity Center

Gravity centers of fish ranges showed seasonal, interannual, and inter climate period variability during 2006–2020 (Figure 6). Fish with a smaller distribution in the study area include P. macrocephalusP. macrocephalus, S. tumbil, and S. undosquamis, while almost all fishes have shown significant banded structure of gravity centers along the northeast-southwest direction. Although one-way ANOVA indicated that the gravity center of each fish species was not significantly different between climate periods (p > 0.05), there was a significant difference between seasons for more than half of the studied fishes (p < 0.05) (e.g., P. anomala, P. macrocephalus, E. cardinalis, A. japonicum, S. undosquamis, U. sulphureus, D. maruadsi, R. kanagurta, and T. japonicus). In addition, the gravity center of most fish species has shifted significantly northward during these 15 years, with only U. sulphureus and moving southward, while the gravity center of U. japonicus and D. maruadsi remained unchanged (Figure 7).

4. Discussion

Population dynamics such as abundance and distribution of fish stocks are closely linked with climate variability, and climate events are undoubtedly the main climatic forcing factors other than seasons [51]. To answer our previous questions, the results of this study showed that: (1) the effect of years on fish abundance appears to be more pronounced than that of climatic events; (2) changes in Chl-a, SSS, and SST were closely correlated to climate events, and abundances of some fish species significantly increased after climate events, especially the fishes of the Carangidae family; and (3) climate events did not affect gravity centers of fish distribution, but most fish species shifted northward over the study period.

The RF analysis of this study identified variables influencing the abundance of each fish species. It seemed that Year, SSS, SST, MLD, and Chl-a were the important predictive variables, but climate events and seasons were of less predictive than years on the temporal scales, suggesting that recent fishing pressure may have caused much disturbance to the fish stocks than climate variability. This is consistent with the previous studies of fishery stocks in the Beibu Gulf, e.g., Su et al. [52] analyzed trophic level and fishing-in-balance index and concluded that fishery stocks in the Beibu Gulf have been overfished with a general downward trend in total fish stock density over the past two decades.

The response of fish abundance to Year in the GAM model corroborated the argument that increasing fishing pressure with the year may be the main factor resulting in decrease in the abundance of most major commercial fishes (e.g., P. anomala, T. lepturus, and T. japonicus). It is also important to highlight that abundances of a few fish species (e.g., S. tumbil, N. bathybius, and R. kanagurta) did not decline obviously with year, which can be explained by the fact that low stock densities may make it difficult to observe the trends, while the decrease in fish abundance also indirectly reduces the ecological competition of the fishes in the same ecological niche and maintains the abundance of the latter under fishing pressure [53]. Despite a series of fishery management and conservation measures that have been implemented by Chinese governmental agencies in the Beibu Gulf since 1999 (including a summer moratorium on marine fishing, a zero-growth policy, the introduction of marine protected areas, and marine ranching), the loopholes in the management measures and the increased fishing pressure from Vietnam have caused decreases in the fishery stocks in this region [54,55]. This trend was exacerbated by the prevalence of unidentifiable low-value and juvenile mixed catch [45,56], which ultimately resulted in the dominant species been replaced by low-value, small-size, and low-trophic-level species [36]. It was suggested that fast-growing populations in variable environments were especially sensitive to overfishing and were more possible to collapse [57]. With a projected global increase in frequency and intensity of extreme climatic events, fish communities in the Beibu Gulf may not only have to tolerate increasing coastal anthropogenic activities but also must adapt to large-scale climatic changes, which would lead to threat of collapse of the entire ecosystem ln the Beibu Gulf.

We also found that high abundances of most fishes occurred between 30–50 m of MLD in the Beibu Gulf, but MLD was not correlated with ONI. Although ocean mixed layer is commonly considered as the column near its surface with vertically quasi-uniform oceanic tracers (temperature, salinity, and density) above a layer of more rapid vertical changes (Lorbacher et al., 2006), its variation in the Beibu Gulf seems to be constrained by more complex factors and unable to reflect the effect of climate events directly. Unlike open oceans, the MLD in the Beibu Gulf, with offshore characteristics, did not exhibit very large spatial and temporal variability [58]. However, as noted above, there is also no evidence that MLD would be significantly affected by climate events, suggesting MLD alone is insufficient to predict how climate events affect fish abundance.

Higher absolute values of ONI, which represent the intensity of climate events, were associated with higher values of detrended SSS and abundances of most fishes increased with SSS. It is well known that salinity is one of the main physical properties that govern the distribution of fish, and numerous studies have revealed that salinity provides the most efficient explanation to fish distribution patterns [59]. Variation of SSS may also cause a series of physiological impacts to marine organisms, such as imbalance of electrolytes, serum hormone levels, and energy metabolism [60]. There are unquestionable evidences that interannual variability of SSS in the South China Sea is also influenced by the ENSO events [61]. Rijnsdorp et al. [62] suggested that while the underlying mechanisms of climate change effect on fish populations remain uncertain, climate-related changes in recruitment is the key process, such as the survival in the pelagic egg or larval stage, or the changes in the quality and/or quantity of nursery habitats. Taking small pelagic fishes (D. maruadsi, T. japonicus, and R. kanagurta) of the Beibu Gulf as an example, the spawning season of those fish species in tropical is mainly in spring and summer [45,63,64], so winter may be the key period for the survival and growth of these species. As indicated previously, salinity is recognized as a key factor that impacts the growth and development of fish, including affecting fish metabolism and survival of eggs and juveniles through osmolarity [65,66]. Past fishing experience shows that Scombrid and Carangid (including D. maruadsi, T. japonicus, and R. kanagurta) pelagic fish are usually considered intermingled fishery stock in the South China Sea [67]. Previous work has also shown that distribution of juvenile and adult D. maruadsi in the inshore waters of China is usually located in high salinity waters [68]. Therefore, increased SSS by climate events may provide a more suitable habitat for these fishes in winter and reduce the pressures to survive. The surplus energy is not channeled into reproduction at an earlier age and/or smaller size [69], thereby reducing the risk of miniaturization and early maturation and indirectly increasing the fecundity of pelagic fish populations [70,71], and ultimately lead to higher abundance in the summer.

Differences found in responses of their abundances to SST in various fishes are probably related to different life history types of the fishes. It is widely known that growth of tropical fishes will show seasonal oscillation and even though only 2 °C temperature differences between summer and winter seasons are adequate to cause a detectable impact [72]. Although changes in temperature have an important effect on the physiological state and behavior of fish [73], we suggest that the intensity of climate events is generally the highest in winter making temperature anomalies more likely to affect primary productivity. Taking the U. sulphureus and P. macracanthus as an example, the results of our study shown that these demersal fishes in the Beibu Gulf have a maximum and a minimum in its abundance in winter and summer, respectively, and these variations also appear to coincide with seasonal variation of Chl-a in the Beibu Gulf. The Chl-a is a convenient biomass proxy for phytoplankton due to its uniqueness to plants and ease of quantification [74], so higher concentration of Chl-a in the winter in the Beibu Gulf indirectly reflects the emergence of phytoplankton blooms, which can sink down and export more organic energy to benthic secondary production. However, because of the degradation and consumption of organic matter during sinking, past studies have indicated that the quality and quantity of food reaching the seafloor are negatively related to water column depth, and lower food availability at depth may in turn affects trophic structure and result in depth-related patterns in biomass and body size in demersal fish communities [75]. However, colder water would obscure the early blooms, which would reduce consumption by pelagic heterotrophs and result in the input of a greater proportion of planktonic production to the bottom sediments than in years when the bloom is later and occurs in warmer water [76]. In addition, the areas with high Chl-a are conducive to zooplankton breeding and gathering, and also attract fish to come for feeding. Although fish may growth faster in the warm periods, the lower Chl-a during this period means that less prey was available for it [77,78], and warming temperatures may compound these challenges by increasing anabolic oxygen demand while decreasing oxygen solubility [79].

Among variations in oceanographic features that are observed following ENSO events are changes in surface seawater temperatures, changes in vertical and thermal structure of the ocean (particularly in coastal regions) [44], and the altered coastal and upwelling currents [80,81]. However, the gravity centers of distributions of the studied fish did not display difference between climate periods. These results indicated that it is difficult to observe significant differences in the analysis of fish distribution and large-scale climate events on small scales spatially. Instead, a significant correlation was found between seasons and gravity centers of half number of the fish species. These may be related to the coastal current from the Qiongzhou Strait interacting with the warm and high saline seawater along the western coast of Hainan Island [82]. There is no doubt that the confrontation varies with seasons under the superimposed influences of the northeast or southwest monsoons. This may explain why the gravity centers of these fishes were distributed along the west coast of Hainan Island. In addition, these results also confirm that spatial distribution of fishes can be strongly influenced by seasonal hydrologic conditions even in structurally uniform environments [83].

It is also noteworthy that gravity centers of most fish species in the Beibu Gulf have shifted northward during the studied period. Even though Beibu Gulf has become a more complex aquatic environment due to damaging effects of anthropogenic inputs, urbanization development, and an expanding industrial marine-economy on the marine ecosystem [84], it seems that these negative conditions do not prevent the gravity center of fish from moving closer to the shoreline for avoiding warmer habitats. Compared with climate events, these results demonstrated that global warming, which has a greater effect on spatial and temporal scales, is the key factor affecting the distribution of fish populations in the Beibu Gulf. Previous studies have shown that ENSO events will seemingly change the distributions of fish, both in the ocean and offshore, and even in coral reefs [85,86]. However, it appears that distribution of habitats on a smaller scale limits the capacity of fish distribution to respond to climate variability. On the contrary, climate change alters the distribution of suitable habitat, forcing organisms to shift their range or attempt to survive under suboptimal conditions [87]. For species with northerly or southerly range margins in the North Sea, half have shown boundary shifts with warming, and most shifted northward [88].

In the context of global warming and decline of offshore fishery resources in China, understanding the influence of global-scale ENSO events on abundance and distribution of commercial fish species is critical for disentangling effects of natural and anthropogenic drivers of change [89]. Such distinctions allow a systematic assessment of nature influences on the fishery stocks and are the basis of informed ecosystem-based management decisions. The predominantly negative effect that years have on fish abundance demonstrates that fishing pressure has a strong effect on fish on long time scales and emphasizes the difficulty of climate events in explaining the change in the overall trend. However, in the short term, climate events will undoubtedly significantly alter environmental variables, such as SSS, SST, and Chl-a. To a certain extent, these environmental variables will also alter the magnitude of seasonal fluctuations in fish abundance, i.e., seasonal peaks or troughs in fish abundance. Our findings did not corroborate the association between climate events and distribution of fishes. It appears that climatic events have less of an impact on distribution of fishes in small areas than seasonal changes. However, on longer time scales, distribution of fish stocks in the Beibu Gulf has shifted northward as a whole due to climate change. Given that China’s offshore fishery management measures and/or policies need to respond to the integrated challenges arising from climate change and anthropogenic activities in the future, further studies should consider the time-lagged effects on commercial fish stocks caused by climate events in different types and intensity, and consequently these effects over fishery ecosystem.