Identification of Keystone Species in Ecological Communities in the East China Sea

Abstract

Keystone species are critical to preserving community stability and species diversity. Identifying key species and investigating their ecological regulation will help to prioritize important species and gain a better understanding of community stability mechanisms. It is based on this understanding that the present study tried to elucidate major keystone species in an important ecosystem in eastern Chinese waters. Therefore, data were collected from fisheries surveys conducted between 2016 and 2021 in the East China Sea. We identified Muraenesox cinereus, Leptochela gracilis, and Trichiurus lepturus as keystone species in the region based on the results of principal component analysis of ten network indices. The removal analysis performed suggested that the loss of keystone species might have a negative impact on the complexity and stability of the food web in the East China Sea. As a result, keystone species should be prioritized in marine ecosystems.

1. Introduction

Keystone species play a decisive role in the structure and function of ecosystems [1]. The concept of keystone species has undergone great evolution since the 1960s. In earlier studies, keystone species were generally considered top predators [2,3,4]. The concept of keystone species was controversial amongst ecologists [5]. Power et al. [6] defined keystone species as those species whose existence has a disproportionately large effect on an ecosystem relative to its population size within that ecosystem. Later, the concept was widely accepted. Species in a community are directly or indirectly related to interactions at different temporal and spatial scales. The extinction of a single species may have a ripple effect on the community, causing more species to be affected or eliminated. However, the extinction of keystone species has an even larger negative ripple effect [7,8,9]. Therefore, the regulation of keystone species plays an important role in maintaining the stability of community structure and interspecific balance [10]. Thus, identifying and protecting keystone species can help to maintain community interactions and improve biodiversity. Marine ecosystems are undergoing major changes owing to climate change, overexploitation, habitat loss, and pollution. These changes have accelerated the decline in ecosystem functioning [11,12,13]. Finding effective measures to reduce biodiversity loss while maintaining ecosystem stability has become an urgent priority for ecologists and fishery managers. In this context, key species have rapidly become a research hotspot in studies on ecosystem dynamics and biodiversity in the field of fishery ecology [14,15,16,17]. Accurately identifying potential keystone species and exploring their ecological regulation mechanisms in marine ecosystems can not only help to determine the priority of species conservation but also evaluate the impacts of certain species on community structure, thus providing a scientific basis for future ecosystem restoration and fishery resource management.

The screening and identification of keystone species is a popular topic in ecological research [18,19,20]. They associate with other species in the food web through interspecific feeding because they are closely related within the food web [21]. Based on this attribute, a quantitative description of the closeness of connections between species and the impact of each species in the food web is crucial for identifying keystone species [18]. In recent years, topology-based ecological network analysis (ENA) has proven to be an effective method for screening keystone species. This can be achieved by calculating a series of topological network indices to quantitatively describe the interactions and influences of various species in the food web [22,23,24]. This method is based on the topological structure of the food web; therefore, qualitative and quantitative feeding relationships are the basis for screening keystone species.

The East China Sea is affected by various water masses, has excellent hydrological conditions, and is rich in food resources. It is a core location for many economically important species to inhabit and reproduce [25]. Wang et al. [26] identified the keystone species of the fish community in the East China Sea and found that Trichiurus lepturus was the keystone predator, whereas Benthosema pterotum was the keystone prey. However, screening and identifying keystone species in the entire community have not yet been carried out. In recent years, fishery resources in the East China Sea have been continuously declining because of overfishing, water pollution, and habitat destruction. The overall community structure shows trends toward miniaturization and a younger age [27]. Identification of keystone species in this community, beyond only fish species, could provide an ecological basis for understanding these recent trends and addressing potential solutions.

In this study, we constructed a qualitative food web based on diet information from the area. In this food web, we ranked the importance of each species in the community to identify the keystone species and then simulated the extinction of species owing to survival pressure. We studied this food web to reveal (1) which species perform the most important functions in the community, (2) how species centrality changes after the removal of one species, and (3) how the structural features of the community as a whole change when important species are removed. This is the first study to use topological network analysis to identify keystone species in the entire community of the East China Sea. These results will help us to understand the food web’s robustness and evolutionary trends in the food web structure. This study provides an important scientific basis for conserving aquatic life and maintaining the structural stability of fish ecosystems.

2. Materials and Methods

2.1. Data

In this study, we screened species to construct a food web based on bottom trawl surveys conducted in the East China Sea during March, May, August, and November of 2016 to 2021. The survey period for each voyage was 7 to 12 days. According to the survey plan, 44 survey stations were completed for each voyage (Figure 1). Sampling was conducted using a bottom trawl with a 20 mm mesh cod end and a trawl mouth opening with a width of 40 m and height of 7.5 m. The bottom trawl was towed once at a speed of 2.2–5.6 kn for 1 h during each sampling process at each survey station.

The types of species were referred to by serial numbers with different initials, i.e., a—fish, b—cephalopods, c—crustaceans, d—other invertebrates and autotrophs. Dietary information was obtained from publications (Table S1) and the FishBase dataset (http://www.fishbase.org/, accessed on 10 April 2022). Based on the diet information of the predators, we constructed a binary matrix of trophic interactions among species within the food web, where rows represented prey species and columns represented predator species. For each cell ij, ‘1’ represented the presence of prey i in the diet of predator j, and ‘0’ represented the absence of this prey.

2.2. Topological Analysis

2.2.1. Global Structural Indices

The global structural indices used were as follows: links (L) represent the number of interactions among species in a food web; link density (LD) is the average number of links of all species and is calculated as $L/S$; connectance (C) is the proportion of actual links to all possible links in the food web and is calculated as $L/S^2$, with higher connectance values indicating increased robustness of the food web [7]; average path length (APL) is the average number of steps along the shortest path for all possible pairs of species in a food web, with a decrease in APL possibly indicating a faster spread of disturbances in the food web [28]; and clustering coefficient (CC) describes the degree of node clustering in the food web [29].

2.2.2. Centrality Indices

The structural positions of the species were characterized using centrality indices. The centrality indices used in this study were degree (D, including in-degree, D_in, and out-degree, D_out), betweenness centrality (BC), and closeness centrality (CL).

The degree of species i (D_i) is the number of other nodes connected to node i [30]. Species that show a high value for D_i are hubs (i.e., connected to many other species). The degree of node i (D_i) is the sum of its prey (in-degree, D_in,i) and predators (out-degree, D_out,i). D_i was calculated as follows:

$$D_i=D_{in,i}+D_{out,i}$$

Closeness centrality (CL_i) quantifies the minimum number of steps from one node to all other nodes [31]. High CL_i values identify species that spread their impact more rapidly to other species when disturbed. CL_i was calculated as follows:

$$C{L}_i=\frac{\left(S-1\right)}{{\Sigma }_{j=1}^S{d}_{ij}}$$

where S represents the total number of species in the food web, and d_ij is the length of the shortest path from species i to j.

Betweenness centrality (BC_i) indicates the frequency with which a species is on the shortest path for each pair of other species. The shortest paths from the bottom to the top were considered in this calculation. The higher the BC_i value, the greater the ability of a species to control information exchange and energy flow in the food web [30]. BC_i was calculated as follows:

$$B{C}_i=\frac{2\times {\Sigma }_{j\le k}{g}_{jk}\left(i\right)/g_{jk}}{\left(S-1\right)\left(S-2\right)}$$

where g_jk is the number of shortest steps between species j and k, g_jk(i) is the number of steps in which species i is present, and S is the total number of species in the food web.

2.2.3. Topological Keystone Index

The keystone index (K) quantitatively describes the importance of a species in a food web [32]. The keystone index consists of both bottom-up and top-down effects. K provides information not only on the number of neighbors of a species but also on how these neighbors are connected to other species. This indicates that K has both direct and indirect effects on the food web. The keystone index of species i was calculated as follows:

$$K_{bi}={\sum }_{c=1}^n\frac{1+K_{bc}}{d_c}$$

$$K_{ti}={\sum }_{e=1}^m\frac{1+K_{te}}{f_e}$$

$$K_i=K_{bi}+K_{ti}$$

where n is the number of predators consuming species i, d_c is the number of prey of the cth predator, K_bc is the bottom-up keystone index of the cth predator, K_bi is the bottom-up keystone index of species i, m is the number of preys eaten by species i, f_e is the number of predators of the eth prey, K_te is the top-down keystone index of the eth prey, and K_ti is the top-down keystone index of species i.

2.2.4. Topological Importance Index

The topological importance index (TI) describes the indirect effects of species through trophic links [19]. TI considers indirect chain effects, and a high TI value indicates a strong ability to diffuse information. The indirect chain effects of species weaken as the number of steps increase [33]. Based on the complexity of the food web in the East China Sea, TI1 and TI19 were selected for subsequent analysis. It was calculated as follows:

$${TI}_i^n=\frac{{\sum }_{m=1}^n{\sum }_{j=1}^S{a}_{m,ji}}{n}$$

(1)

where a_m,ji is the effect of i on j when i reaches j across m steps, and ${TI}_i^n$ is the topological importance index that describes the effect of species i on the food web topological structure across n steps.

All indices were integrated into a comprehensive index using factor analysis with the principal component (PC) method after standardization [34]. Factor analysis can be used to transform the multidimensional space of multiple indices into a reduced space of PCs through appropriate mathematical transformations and to select a few PCs that account for a large proportion of the total amount of information in the variation as a new PC. The principal component analysis (PCA) transforms the original indicators into independent PCs. This new variable is useful for characterizing the positional importance of the species [35].

2.2.5. Removal Analysis

Species were removed based on the ranking of the PCA results, and the effects of their removal on other species were assessed based on changes in their centrality indices. To quantify the effects of species removal, we calculated changes in three centrality indices: D, BC, and CL. The effects of species removal on D, BC, and CL were calculated as follows:

$${△A}_i=\left|A_i^{rem}-A_i\right|$$

where $A_i^{rem}$ represents the centrality indices after species removal, and Ai represents the centrality indices in the original food web. Therefore, a value of zero indicated no change. The larger the value, the larger the centrality of the change after removal.

We also calculated global structural indices (L, LD, C, APL, CC) to examine the effects of the lack of species with high topological status on the structure of the food web. In addition, we simulated the cumulative removal of species and calculated the change in global structural indices with an increase in the number of species removed to test the robustness of the food structure web against species loss.

A food web diagram was plotted using NetDraw in UCINET (ver. 6). The centrality indices were calculated using UCINET (ver. 6), whereas K and TI were calculated using CoSBiLab Graph (http://www.cosbi.eu/, accessed on 19 April 2022). The removal analysis was performed using R software (https://www.r-project.org/, accessed on 2 May 2022).

3. Results

3.1. Topological Analysis

In the present study, the food web of the East China Sea consisted of 58 single species and 18 trophospecies with 730 feeding relationships. The global structural indices are listed in

Table 1. Global structural indices of the food web in the East China Sea.

Indices	S	L	LD	C	APL	CC
Value	76	730	9.605	0.126	1.128	0.174

. The food web is shown in Figure 2.

All network indices differed significantly between species and trophospecies, except CL (P_D = 0.019; P_Din = 0.000; P_Dout = 0.000; P_CL = 0.068; P_BC = 0.000; P_TI1 = 0.000; P_TI19 = 0.000; P_k = 0.009; P_kt = 0; P_kb = 0.021). This suggested that the division of species and trophospecies will have a considerable effect on the calculation of the indices because trophospecies consist of many species. Therefore, we considered only species in the subsequent analyses.

In the East China Sea, top predators such as Scoliodon laticaudus, Ilisha elongate, Lophius litulon, Saurida tumbil, and Miichthys miiuy were at the top of the food web (D_out = 0). Apogonichthys lineatus, Johnius belangerii, Pampus argenteus, Psenopsis anomala, Solenocera melantho, Parapenaeopsis hardwickii, and Metapenaeopsis barbata were found in the middle of the food web. The D, CL, and BC values for T. lepturus, Muraenesox cinereus, A. lineatus, Leptochela gracilis, Larimichthys polyactis, and Oratosquilla oratoria were the highest, indicating that these species had more trophic interactions in the food web and could transmit information to other species at a fastest rate. They were more likely to affect the transfer of information and energy among other species in the food web.

The predators with higher K_t values, such as fish species S. laticaudus, M. cinereus, L. litulon, S. tumbil, I. elongate, and T. lepturus, had greater top-down effects on the food web. Invertebrates with higher K_b values, such as L. gracilis, O. oratoria, and Alpheus japonicus, showed a greater downward effect on the food web. The species with the top 20 K values were all fish species, except for L. gracilis and O. oratoria, indicating that fish species dominated the food web.

Based on the analyses of TI1 and TI19, the species L. gracilis, O. oratoria, A. lineatus, Palaemon gravieri, Chaeturichthys stigmatias, and B. pterotum were found to have high TI1 and TI19 values, indicating that they could spread the most information through the food web in the shortest time, either directly or indirectly.

3.2. Principal Component Analysis

To evaluate the importance of species in the food web, we performed PCA on the ten topological indices of the East China Sea. This method allowed us to simplify the multidimensional data structure of the topological network indicators. A comprehensive index for evaluating the keystone species was calculated using two independent principal components, which explained 88.759% of the variance.

PC = 0.3199 × D + 0.2529 × D_in + 0.1415 × D_out + 0.3135 × CL + 0.3016 × BC + 0.2136 × K + 0.1352 × K_b + 0.1866 × K_t + 0.1412 × TI1 + 0.1331 × TI19

According to the PCA, M. cinereus, L. gracilis, and T. lepturus were ranked in the top three, comprehensively. Combined with the topological indices ranking, these three species ranked first in eight indices (

Table 2. Network indices of food web in the East China Sea.

D		Din		Dout		CL		BC		K		Kb		Kt		TI1		TI19		PC
a10	39	a02	35	c17	31	a10	67.568	a02	4.937	a01	26.23	c17	4.29	a01	26.23	c17	2.62	c17	0.23	a02	4.544
a02	36	a10	32	c06	21	a02	65.789	a10	3.594	a02	17.17	c06	2.13	a02	17.11	c06	1.54	c06	0.11	c17	4.16
c17	33	a06	24	a22	15	c17	64.103	c17	2.569	a04	13.59	c05	1.39	a04	13.59	a22	1.01	a22	0.07	a10	3.881
a22	33	a03	24	a26	15	a22	64.103	a22	2.259	a08	10.69	a22	1.30	a08	10.69	c07	0.95	c07	0.07	c06	2.651
a18	32	a20	23	c07	13	a18	63.559	a18	2.195	a03	10.36	a26	1.30	a03	10.36	a28	0.91	a28	0.07	a22	2.58
c06	31	a18	21	a28	12	c06	63.025	c06	2.135	a10	8.89	a28	1.29	a10	8.46	a26	0.90	a26	0.07	a18	2.338
a26	29	a14	20	a18	11	a26	61.983	c08	1.403	a06	5.76	c07	1.27	a06	5.61	c05	0.84	c05	0.07	a01	2.013
a20	28	a01	20	a24	11	a20	61.475	a06	1.381	a05	4.89	c15	1.27	a05	4.83	a18	0.81	c15	0.07	a26	1.613
a06	27	a05	19	a33	11	a06	60.976	c07	1.312	c17	4.37	a24	1.01	a09	3.52	a24	0.81	a18	0.05	a06	1.338
a23	26	a22	18	a31	11	a23	60.484	a14	1.306	a07	3.73	a33	1.01	a07	3.50	c15	0.78	a24	0.05	a20	1.178
a24	24	a04	17	c15	11	a03	59.524	a24	1.211	a09	3.58	a18	0.96	a13	3.46	a33	0.75	a33	0.05	a03	1.13
a03	24	a08	17	a23	10	c07	59.055	a20	1.123	a13	3.46	c09	0.95	a20	3.15	c09	0.73	c09	0.05	c07	1.028
c07	23	a23	16	c05	10	a24	59.055	a26	1.026	a20	3.42	c13	0.93	a14	3.10	c13	0.65	c13	0.05	a24	1.027
c08	22	a07	16	c09	10	c08	58.594	a23	1.019	a18	3.21	a31	0.83	a15	2.95	a31	0.64	a31	0.04	a23	0.999
a21	22	a19	15	c08	9	a21	58.594	a28	0.815	a14	3.17	a23	0.73	a19	2.52	c11	0.59	c11	0.04	c05	0.663
a14	22	a26	14	a21	8	a14	58.594	a17	0.802	a15	2.95	c11	0.73	a16	2.26	a23	0.56	a23	0.04	c08	0.573
c05	21	a21	14	a25	8	c05	58.140	c05	0.787	a22	2.83	a21	0.57	a18	2.24	c08	0.53	c08	0.03	a28	0.513
a07	21	a17	14	c11	8	a07	57.692	a03	0.776	a24	2.81	c08	0.56	a12	2.06	a21	0.46	a21	0.03	a14	0.459
a28	20	a24	13	a32	8	a01	57.692	c09	0.760	a19	2.69	a25	0.56	a24	1.8	a32	0.43	a32	0.03	a04	0.386
a05	20	c08	13	a10	7	a28	57.252	a08	0.709	c06	2.66	a32	0.51	a11	1.61	a17	0.42	a25	0.03	a21	0.325

). Therefore, the three species were identified as keystone species.

3.3. Removal Analysis

Based on the ranking importance of the PCA, we simulated the removal of the first 30 species. The effect of removal on a single species, assessed by computing the centrality indices, is presented in Figure 3, Figure 4 and Figure 5. The results showed that the number of species with varying D values decreased when the species ranking was removed. After the removal of M. cinereus, which had the highest CL value, the greatest changes were observed in the other species. For the three species with the highest BC values (M. cinereus, L. gracilis, T. lepturus), the effects were greater than those following the removal of species with low closeness values.

The global structural indices changed after their removal, as shown in Figure 6. The results showed that the higher the species was ranked, the greater the impact on the global structural indices after removal. The removal of T. lepturus had the greatest influence on the global structure. As an exception, S. laticaudus with its high rank had little effect on the global structure.

After cumulative removal according to the importance ranking, all global structural indices showed a general downward trend with an increase in the number of removed species, and the rate of decline gradually decreased (Figure 7). After removing the first 15 species, the trends in C, APL, and CC tended to stabilize and their values showed slight fluctuations.

4. Discussion

4.1. Complexity of Food Web

Keystone species play a decisive role in the structure and function of ecosystems. Research on keystone species plays important roles in understanding the complexity of food webs and protecting biodiversity. The complexity of the food web was mainly measured by a series of global structural indices extended by the number of species (S) and the number of linkages (L). In our study, food webs in other areas of the China Sea and some international trending areas were selected and compared with the global structural indices of the East China Sea (Table 3). Studies have shown that robustness may be improved with increases in LD and C of food webs [36,37]. Compared with other marine food webs, the LD and C values of the East China Sea were higher. It may be that the community is dominated by omnivorous species of medium trophic level, which tend to interact with more species from more than one trophic level. MacArthur [38] showed that omnivorous species can have more food choices than monophagous species after the extinction of a species. The omnivorous characteristic reduces the risk of species extinction, which gives trophic flexibility to an ecosystem. Dunne et al. [7] showed that the APL values of marine food webs were lower (1.6 links) than those of other types of food webs. The APL value of this food web was significantly lower than that of other marine food webs, suggesting that negative effects could spread rapidly and widely throughout the community. In addition, the CC value of this food web was average compared to that of other food webs. This suggested that species were closely related to each other and likely to cluster around keystone species [29].

It is worth mentioning that the relationship between the complexity of food webs and biological invasion has long been discussed. Invasive species can affect the function of ecosystem by reducing population abundance, decreasing native biodiversity, and squeezing the living space of native species [39]. Many studies have shown that the complexity of native food webs may be an important factor in determining invasion success, and connectance is widely used to quantify the resistance of food webs to invasions [40,41,42]. There are two different views on the relationship between connectance and invasibility. One view is that high-connectance food webs may in some cases lead to low invasibility [40,42], while the other view suggests the opposite [41]. There have been no previous reports of invasive species in this sea area. This may be due to the fact that invasive species are more likely to encounter generalist enemies in high-connectance food webs [43]. In addition, several keystone species in the native food web of the sea area are better adapted to competition and predation, and thus the food web is more resistant to the negative effects of invasive species [44].

Table 3. Global structural indices of food webs in different sea areas.

FOOD WEB	LD	C	APL	CC
East China Sea	9.6	0.13	1.13	0.174
Yellow Sea [45]	4.4	0.04	NA	0.138
Laizhou Bay [17]	3.0	0.18	NA	0.325
Haizhou Bay [46]	11.0	0.12	2.11	0.230
the sea adjacent to Miaodao Archipelago [47]	6.0	0.16	2.10	0.301
Arctic [48]	8.6	0.05	2.28	0.250
Antarctic [49]	6.8	0.01	3.00	0.140
Potter Cove [29]	3.4	0.04	1.80	0.080
Caribbean coral reefs [50]	11.1	0.22	1.60	0.360
Golfo de Tortugas [1]	2.5	0.01	1.65	0.022

Table 3. Global structural indices of food webs in different sea areas.

FOOD WEB	LD	C	APL	CC
East China Sea	9.6	0.13	1.13	0.174
Yellow Sea [45]	4.4	0.04	NA	0.138
Laizhou Bay [17]	3.0	0.18	NA	0.325
Haizhou Bay [46]	11.0	0.12	2.11	0.230
the sea adjacent to Miaodao Archipelago [47]	6.0	0.16	2.10	0.301
Arctic [48]	8.6	0.05	2.28	0.250
Antarctic [49]	6.8	0.01	3.00	0.140
Potter Cove [29]	3.4	0.04	1.80	0.080
Caribbean coral reefs [50]	11.1	0.22	1.60	0.360
Golfo de Tortugas [1]	2.5	0.01	1.65	0.022

4.2. Keystone Species

According to the PCA and topological indices ranking, M. cinereus, L. gracilis, and T. lepturus were identified as keystone species. The D, D_in, BC, CL, K, and K_t values of M. cinereus were ranked first and second, respectively, indicating that M. cinereus had more feeding relationships, stronger top-down effects, and stronger information transmission abilities in the food web. M. cinereus is a highly trophic fish that feeds on both benthic and swimming organisms. It mainly feeds on secondary consumers, such as O. oratoria, L. polyactis, and some shrimp. In the food web, M. cinereus had the highest number of prey species at 35. The D_out, K_b, TI1, and TI19 values of L. gracilis were ranked first, showing that this important food organism had important bottom-up effects and a strong information diffusion ability throughout the food web. The D, D_in, CL, and BC values of T. lepturus were ranked first or second, respectively. T. lepturus is a swimming species that feeds on benthic animals and zooplankton. The main feeding groups are fish, krill, stomatopods, and cephalopods. T. lepturus is a fierce predator with diverse feeding habits, similar to M. cinereus, and it occupies similar positions and performed similar functions in the food web.

M. cinereus and T. lepturus were the top two species in terms of centrality indices. These centrality indices suggested that M. cinereus and T. lepturus played important roles in the spread of direct and indirect effects through the ecosystem via the shortest path. Most energy in the food web travels along the shortest path [51]. Control can be achieved from higher to lower trophic levels via several food web pathways. The fastest transmission paths are ensured by the shortest paths, which tend to be the main routes through which top predators can achieve top-down effects [52]. In terms of BC, M. cinereus and T. lepturus were likely key links in the direct interactions between high and low trophic levels.

4.3. Removal Analysis

As shown in Figure 3, Figure 4 and Figure 5, the removal of M. cinereus caused the greatest changes in the centrality of other species. The CL and BC values of many species decreased after the removal of M. cinereus, indicating that a decline in M. cinereus may lead to longer energy transfer paths in the food web, thus losing transfer efficiency and leading to the reorganization of energy flow. However, in a real food web, species actively respond to extinction through mechanisms such as feeding shifts to cushion these effects [53]. Despite this, reorganization of energy flows in the ecosystem could still occur, affecting the total energy of top predators and perhaps even fisheries in the area [54].

To quantify the impact of species extinction on the global network, we simulated the removal of the first 30 species in order of importance. The first five species had a significantly greater impact on the global network indices than the bottom species, with T. lepturus having the greatest impact after removal (Figure 6). According to Chen et al. [55], T. lepturus has a strong resource utilization ability and tends to feed locally, maintaining a good nutritional connection with the surrounding prey. T. lepturus has been shown in numerous studies to be the dominant species in the East China Sea, with abundant resources, ensuring its top-down effects in the food web [27,56]. As an omnivorous species, T. lepturus can adjust its predation pressure according to the abundance of prey, thus making the resource utilization of the ecosystem more efficient, which is not possible for monophagous species [57]. Omnivorous species provide ecosystems with a better buffer against environmental disturbances. Ecosystems respond more quickly to disturbances caused by omnivorous species associated with species at different trophic levels. Omnivorous species are usually at the third or higher trophic levels, and if perturbations affect the lower trophic level, omnivorous species directly related to this level can respond quickly [58]. Monophagous species must wait until the disturbance reaches their trophic level before responding, which may result in a longer response time.

Five global structural indices showed obvious downward trends after cumulative removal (Figure 7). After removing the first 30 species, the L value dropped from 730 to 205 and the LD value dropped from 9.6 to 4.5 among the three indices (L, LD, C), reflecting the complexity of the food web. This suggested that the food web tended to become simpler and less complex after extinction of the top-ranked species and that the rate of decline slowed as the importance of the species declined. In addition, the APL and CC values gradually leveled off after the first 15 species were removed. A decrease in APL means that perturbations spread faster through the food web, which should be buffered by keystone species such as M. cinereus and T. lepturus to reduce the impact of disturbances on the food web. When keystone species are removed, the perturbations affect other monophagous species at a faster rate, causing the food web to become shaken and less stable. The level of CC may depend on the majority of species in the food web rather than on specific species, but omnivorous species greatly affect clusters by feeding across and within trophic levels [29]. These omnivorous species promote energy transfer between habitats during feeding [59]. Therefore, the decline in CC tended to flatten after the removal of the top species.

5. Conclusions

In this study, we constructed a binary network based on interspecific feeding relationships in the East China Sea. The LD, C, and CC values of the food web were high, which indicated that the food web had strong robustness and resistance. The lower APL value meant that negative effects spread rapidly and widely throughout the food web.

M. cinereus, L. gracilis, and T. lepturus were identified as the keystone species in this area according to 10 topological indices. L. gracilis, as a key food organism, plays an important upward effect in the community. Both M. cinereus and T. lepturus serve as key links between high and low trophic levels in the food web. According to the removal analysis, the difference between the two species was that the loss of M. cinereus changed the centrality of other species in the community more, while the loss of T. lepturus had a greater effect on the global structure of the food web.

Based on some inherent limitations of the topology network analysis approach, our research had some shortcomings that need to be improved. Binary networks are unable to quantify the strength of interspecific interactions. This information is crucial to understanding how energy flows in a community. In addition, a topological network, as a static network in the form of snapshot, cannot show the dynamic changes in an ecosystem. These limitations point the direction for our further research in this field.