Analyzing network topological characteristics of eco-industrial parks from the perspective of resilience: A case study

Ecological Indicators 74 (2017) 403–413

Contents lists available at ScienceDirect

Ecological Indicators journal homepage: www.elsevier.com/locate/ecolind

Short communication

Analyzing network topological characteristics of eco-industrial parks from the perspective of resilience: A case study Xiangmei Li a , Renbin Xiao b,∗ a b

School of Economics, Environment and Resources, Hubei University of Economics, Wuhan 430205, China School of Automation, Huazhong University of Science and Technology, Wuhan 430074, China

a r t i c l e

i n f o

Article history: Received 4 February 2016 Received in revised form 19 October 2016 Accepted 21 November 2016 Available online 9 December 2016 Keywords: Eco-industrial parks Industrial ecology Symbiotic network Complex network Resilience Topology

a b s t r a c t Realizing the stable operation of an eco-industrial park (EIP) as a complex system consisting of a variety of the enterprises and embedded relations is challenging. The topological structure plays an important role to understand the balance of network resilience and eco-efficiency in the operation process of a given EIP. In this paper, Ningdong Coal Chemical Eco-industrial Park (Ningdong CCEIP) is used as a case study in Ningxia Hui Autonomous Region of China. Based on complex network theory, we focus on topological characteristics analysis of symbiotic network from the perspective of resilience. Results reveal that Ningdong CCEIP has scale-free characteristics as well as the small world ones. Compared with the node-level metrics, the important degree of node considering ecological factor is a more crucial index measuring the importance of a particular node in the network. The removal of top 10% node contributes to 60% decrease of network efficiency, which indicates the decline of resilience in the studied case. Protecting the most important nodes is critical to safeguard the potential “vulnerability” in the development of EIPs. This study can help us better understand the strategies for avoiding disruptions, improving the resilience of EIP and safeguarding the stable operation. © 2016 Elsevier Ltd. All rights reserved.

1. Introduction The expansion of resources consumption and the aggravation of environmental decay, coupled with increasing urbanization, rising population size and accelerating economic development have imposed a considerable impact on the planet (UNEP, 2000). Tammemagi (1999) contended that a “mere 1.5 ha of space will be available to each individual for housing, food production, waste disposal and other needs” by 2020. In the context, there have been increasing calls for eco-efficiency, aiming at reducing material and energy throughput without influencing goods and services supplied (Gibbs and Krueger, 2005; Zhu and Ruth, 2013). In 1989, the term “industrial ecosystems”, firstly introduced by Frosch and Gallopoulos (1989) in Scientific American, has served as one of the important solutions to achieve productive use of waste and by-products and minimize environmental degradation. Meanwhile, a cluster of companies from different industries, were intensively sharing resources to competitive advantage involving physical exchange of materials, energy, water, and by-products, which was defined as industrial symbiosis (IS) by Chertow (2000).

∗ Corresponding author. E-mail addresses: [email protected] (X. Li), [email protected] (R. Xiao). http://dx.doi.org/10.1016/j.ecolind.2016.11.031 1470-160X/© 2016 Elsevier Ltd. All rights reserved.

A variety of companies and embedded symbiosis relations among the companies forms industrial symbiosis networks (ISNs), which often was expressed as eco-industrial park (EIP) in the research literature (Côté and Cohen-Rosenthal, 1998). EIP has coemerged with the focus on efficiently sharing resources in a similar fashion to natural ecosystem in the pursuit of reduced ecological impact and maximized economic benefits (Liwarska-Bizukojc et al., 2009). In recent years, attention to EIP development projects has been attracting at a good pace both in developing and developed countries worldwide, including USA (Chertow and Lombardi, 2005), Australia (Van Beers et al., 2007), Canada (Venta and Nisbet, 1997), Finland (Korhonen et al., 1999), India (Patel et al., 2001), Korea (Behera et al., 2012), and China (Zhu et al., 2007). Meanwhile, many researchers have been studying ISNs (Wright et al., 2009; Boons et al., 2011; Behera et al. 2012), in order to provide new implications on the design and improvement of EIPs. To develop cost-effective strategies to conserve resources and reduce wastes, Tan et al. (2008) applied systematic approaches to find optimal designs of industrial material reuse/recycle networks. Wright et al. (2009) quantitatively interpreted industrial parks as ecosystems, by translating ecological tools, such as diversity and connection, to an industrial context, and presented a theoretical platform to explore how the ecological concepts affect the sustainability of industrial ecosystems.

404

X. Li, R. Xiao / Ecological Indicators 74 (2017) 403–413

However, the outstanding problems are the instability and poor symbiosis confronted by many national EIP projects (Zhu et al., 2010; Behera et al. 2012), in spite of the fact that EIP can maximize resource efficiency and minimize pollutant emissions (Park and Behera, 2014). Resilience, as a property of complex systems, refers to industrial system’s ability to sustain in an instable market environment in response to unanticipated perturbations (Korhonen and Seager, 2008). Its ecological meaning has been extended to provide guidelines for design and management in the industrial ecosystems increasing firm and system level adaptability to disruptions (Zhu and Ruth, 2013). Among a range of limitations, the structure of the network is very important factor that is influenced by perturbation, and further affects a system’s resilience (Gao et al., 2016). The introduction of network theory opened a new way for many scholars to study the topological structure in the EIS as a complex network (Korhonen and Snä kin, 2005). It was applied on Kalundborg industrial network to understand its organizational framework, i.e. the structural characteristics of industrial symbiosis and the role that different actors play (Domenech and Daviesa, 2011). Based on the intrinsic properties of EIS, Xiao et al. (2012) proposed a modified Baraba´ısi and Albert (BA) model to design the green logistic supplying network to accurately interpret the formation mechanism of an eco-industrial network. To help EIP managers understanding the importance of stability and identify sustainable strategies, several researchers have been working on resilience in the eco-industrial networking under disruptive scenarios (Chopra and Khanna 2014; Zeng et al., 2013; Zhu and Ruth, 2013), to provide effective support for industrial ecology. Zhu and Ruth (2013) proposed the concept of resilience for industrial ecosystems and explored the influencing factors of resilience to removal of firms from networks by employing a network model of inter-firm dependency. Chopra and Khanna (2014) applied extensively network analysis to enhance understanding the resilience based on the network metrics in response to partial (untargeted disruption on a node) and complete (targeted attack) disruptions, from 2002 snapshot and evolution analysis during 1960–2010. Zeng et al. (2013) put forward the critical threshold to quantitatively assess the resilience of EIPs. Xiao et al. (2016) simulated the network cascading failure to show how the node enterprises impact the stability of EIP. However, the available related studies had not provided us an effective assessment of resilience characterizing the integrated feature of EIP. The ecological relationships among members should be gained important insights in EIPs, which results from direct and indirect exchange of byproducts and wastes (Zhang et al., 2015; Zeng et al., 2013). And mutual dependence among nodes (Hu et al., 2015) and node attributes (Chen et al., 2016; Liu et al., 2015) may provide useful information for structural exploration. Therefore, network topological characteristics, with ecological attributes, different from other general complex system, should be considered to study the important node and the resilience of highly interconnected and symbiotic industrial network. In this paper, complex network theory is applied to focus on network topological characteristics from the perspective of resilience. Using a case study of Ningdong Coal Chemical Eco-industrial Park (Ningdong CCEIP), one of EIPs in Ningxia Hui Autonomous Region, this paper has three specific objectives to obtain: (1) to calculate the statistical parameters of EIP and to identify the topological characteristics of the case study; (2) based on node-level metrics and ecological factor, to quantitatively determine the important enterprises by a novel evaluation model; (3) to explore the change of resilience in response to targeted disruptions. The study relies on complex network theory to provide a systematic overall view of the structure and function based on topological structure and ecological features in EIP, which will contribute to further expand the theoretical framework for developing resilience of EIP.

2. Methods 2.1. Research background In this paper, Ningdong CCEIP is used as a case study, which is one of the three industrial parks in Ningdong of Ningxia Hui Autonomous Region in northwestern China. The chemical industrial park combines traditional coal chemical industry (such as coal to methanol) with new coal chemical industry (such as coal to olefin), and produces many kinds of products, such assyngas (SNG), polyacrylonitrile, polyethylene, poly trimethylene terephthalate (PTT), poly ethylene terephthalate (PET), poly oxy methylene (POM), ␥-butyrolactone, tetrahydrofuran, naphtha, diesel oil, coal tar, etc. In particular, polyethylene accounts for the largest amount of synthetic resin, which is mainly applied in hollow products, plastic products, wire, etc. PTT is a kind of new polyester polymer material, which has broad application prospects in many areas. POM is one of the five big engineering plastics in the world and also called ‘super steel’, which can substitute metallic materials (or metals) such as steel, copper, zinc and aluminum in many parts of the industry. SNG is a high-quality and clean energy, which can be used in power generation, chemical raw materials, natural gas cars, etc. Ningdong CCEIP consists of four industrial systems: coal-fired power, coal chemical industry, salt chemical and glass building materials, based on the principles of ‘development plan focusing on the keys, advancing eco-industrial construction step by step, coal-fired electricity industry as the foundation, resources in situ conversion, deep processing and comprehensive utilization system’. Its main chains are as follows: (1) Coalmine → SNG, semi-coke, tar, methanol, coal gangue → all kinds of chemical products (such as naphtha, POM, etc.). (2) Coalmine → thermal power plant → electricity. And at the same time, the thermal power plants provide steam for chemicals. (3) Coalmine → polyvinyl chloride (PVC) → various chemical products (such as ammonia, etc.) and instrument components (such as bellows). In addition, the exchange of ‘by-product to raw material’ can reduce the resources consumption and environmental pollution, and generating considerable economic benefits. The schematic diagram of system network of Ningdong CCEIP is shown as Fig. 1.

2.2. Network metrics Complex network can be considered as an approach to characterize complex systems in the real world (Strogatz, 2001). In a network, nodes identify the elements of the system and the set of connecting links (edges) represent the presence of a relation among those elements (Barrat et al., 2008). Network theory (also called graph theory) can be traced back to 1730s thanks to Leonhard Euler ¨ (1736) in solving the Konigsberg seven bridges problem. In 1998, Watts and Strogatz released a paper named “Collective Dynamics of Small-world Networks” in Nature and established a small world model. Besides, Professor Barabasi and Dr. Albert published a paper named “Emergence of Scaling in Random Networks” in Nature and established a scale-free networks model. So far, complex network has become a hot spot of complexity science, which is widely applied in food web, WWW, electric power grids, railway networks etc. (Strogatz, 2001; Albert and Barabási, 2002; Hong et al., 2015). Because structure always affects function (Xu, 2000), the differences in topology of the network would bring about the diversity of system functions. For instance, the topology of social networks affects the diffusion of information and diseases, and the topology of the power grid affects the robustness and stability of power transmission (Strogatz, 2001). To describe the topological characteristics of network, the paper mainly pay close attention to

X. Li, R. Xiao / Ecological Indicators 74 (2017) 403–413

Coal chemical industry Y-Butyrolactone

POM

BDO

PET

glycol

PTA

Coal gasification

NaOH

SNG

semi coke

coal tar

CO2

semicoke

diesel

H2 O2

NH3

Cl2

H2 CO CH4 CO2

coal mine

DME

Waste gas

CaC2

PVC

NaOH

Na2CO3

SiHCl3

methane oxide organic silicone

coal gangue

waste gas coal cinder

Waste water

waste water cinder

methanol slag

power generation

condensed water

flyash

coal mine Thermal power plant

Waste residue

Sewage treatment plant

river

Lime steam

liquefied gas

Cl2

sulphur slag waste water

HCl NaClO H2SO4

carbide slag

CPVC

coal pitch

Methan ol

polyvinyl

Vinyl

H2

corrugated pipe

upgrading

Coal mine

Ethynyl

acetate N2

Coal mine

naphtha

cinder

urea

Methanol

CO2 Benzene

tar

waste gas

waste water

CO

MTO

liquid ammonia CaC2

Waste gas treatment

1,3Acryloni propane polyethyle trile diol ne

MTA

Solid caustic soda

liquefied gas

polyacrylonitrile

Coal gasification

tetrahydrof uran

PTT

405

CaSO3

cinder

reuse of reclaimed water

CO NOX SO2 Wastewater treatment and recycling

carbide slag

slag flyash

cement thermal insulation material brick

Waste residue treatment

Fig. 1. Brief system network of Ningdong CCEIP.

shortest path length, degree, degree distribution, clustering coefficient, and so on. 2.2.1. Average shortest path length In a network, the shortest path length dij is defined as the number of edges from node i to node j traversed by the shortest connected path. The diameter is traditionally defined with the shortest path length as follows. D = max{d(i, j)}

(1)

In addition, the average shortestpath length is defined as the average value of dij over all the possible pairs of vertices in the network, which indicates the average degree of separation between the node pairs, and could be considered as the measure of the efficiency for information transmission.



L=

dij

i>j 1 N(N 2

− 1)

(2)

Where N denotes the number of nodes in a network.

Fig. 2. The scheme illustrates the correlation between the a node i and its neighboring nodes, i.e., nodes i1, i2, i3, and i4.

2.2.2. Degree and degree distribution Node degree is defined as the number of edges that are connected to a node (or the degree k of node i: ki , shown by Fig. 2). Average degree of a network is then the sum of the node’s degrees divided by the number of nodes that exist in the network N 

k =

i=1

N

ki (3)

The definition given above adapts to undirected graphs perfectly. In directed graphs, it is possible to differentiate the in-degree of a node i (kin,i ) and the out-degree of a node I (kout,i ). The indegree represents the incoming edges connected to node i, while the out-degree measures the number of edges outgoing from node i. The degree of a node (i.e., total degree of a node) in a directed graph is defined by the sum of the in-degree and the out-degree, ki = kin,i + kout,i .

406

X. Li, R. Xiao / Ecological Indicators 74 (2017) 403–413

In the case of directed graphs, two degree distributions p(k), including the in-degree p(kin ) and out-degree p(kout ) distributions are defined as the probability that a randomized node has in-degree kin and out-degree kout , respectively. Specifically, degree distribution of random network approaches a Poisson distribution for large N (Albert and Barabási, 2002), and that of small-world networks is rather homogeneous to that of random network, while degree distribution of scale-free networks follows the power-law distribution. 2.2.3. Clustering coefficient Given node i, the clustering Ci of node i is defined as the ratio of the number of links among the neighbors of i and the maximum number of such links. If the degree of node i is ki and if these nodes have fi edges among them, there is Ci =

fi ki (ki − 1)/2

(4)

 ij (v) i= / v= / j

ij

(5)

Node betweenness represents the effects of the corresponding nodes across network, which is significant to explore and protect the key resources and technology in actual network.

Re =

ka

ua ∈ U2

(7)

N(N − 1)/2

Where ka is the degree of node a, and



ka is the summation

degree of ecological nodes. The higher the value of Re is, the more the flow of the reciprocal utilization of by-products and wastes within EIP is. EIPs exchange waste and by-products through interactions between companies and through integrated resource recovery systems in order to maximize resource efficiency while minimizing pollutant emissions (Lowe and Koenig, 2006). The disruption on a node, like shortage of a resource, minor technical failure, etc. would cause an impact and make the whole system less eco-efficient. Therefore, comparing the value of the rate of ecological connection before and after disruption is more important for determining the critical node. Here, we propose another indicator, i.e. ecological effect of node, to characterize the node’s impact on eco-efficiency in the EIPs.



Ii = 1 −

2.3. Resilience analysis of the EIP In order to evaluate the resilience of the Ningdong CCEIP, we calculate the change of network efficiency in respond to the disruption of important nodes. The resilience analysis of EIPs includes following two sections: (1) proposing a novel evaluation model for selecting the important nodes; (2) when considering the scenario of disruption induced by removing the important node i, analyzing the change trend of network efficiency based on the load and capacity on a node. In actual complex networks, evaluating the node importance need multiple-sided ways, not only topological characteristics such as node degree and node betweeness, but also node attributes such as useful information (Chen et al., 2016). EIPs are based directly on industrial ecology principles theoretically minicking the operation of natural ecosystems (Liwarska-Bizukojc et al., 2009). Ecological functions (attributes) of nodes played an important role in the operation of EIP. Therefore, we construct important degree of node (ID) by integrating topological indicator and ecological factor indicator in order to evaluate the node importance of EIP in our paper. Fig. 3 shows the constructing flow chart of ID. The novel evaluation model of the important nodes is as follows. IDi = i ∗ INC = i ∗ (˛1 ki + ˛2 Bi )



ua ∈ U2

2.2.4. Betweenness Node betweenness and edge betweenness are often introduced to measure the importance of a node in a network. In this paper, attention is paid to node betweenness. It is defined as the number of shortest paths between pairs of nodes that pass through a given node. More precisely, if ij is the total number of shortest paths from i to j, ij (v) is the number of these shortest paths that pass through node v, the betweenness of v is defined as B(v) =

structure. Therefore, a1 = a2 = 0.5. ki , Bi is the normalized value of ki and Bi respectively. The detailed description about Ii is as follows. Supposing the set U of nodes in the networks  is divided into two categories U = U1 ∪ U2 , where the set U2 = m1 , m2 , ..., mNp consists of waste nodes and by-product nodes which are all called ecological nodes, and the U1 = {u1 , u2 , ..., uNe }consists of all other nodes which are called product nodes (or un-ecological nodes). The number of all nodes is N = Ne + Np . We propose a parameter which is called the rate of ecological connection Re . It is defined as follows.

(6)

where IDi is the important degree of node i, i is the ecological factor of node i, which is equal to Ii /Imax obtained by ecological effect of node i (Ii , the detailed description about Ii as follows) and Imax refer to the maximum of Ii , and INC represents integrated node centrality. a1 , a2 represent the weight of node degree (ki ) and node betweenness (Bi ). In the paper, we suppose the contributions of ki and Bi are the same for evaluating important nodes from topological

AR =1− BR

kb i

ub ∈ U2 ∩U



(8)

ka

ua ∈ U2

Where Ii is ecological effect of node i. BR and AR are the rates of ecological connection before and after disruptions in EIPs, respec tively. Ui is the set of residual nodes under the normal operations in the whole network after disruption induced by removing node i. That is, the numerator of the fraction is the summation over the degree of all residual ecological nodes after disruption caused by the removal node, and the denominator of the summation over the degree of all ecological nodes before disruptions. The higher the value of indicator Ii , the greater the impact of node i on the ecoefficiency of EIP. In this paper, important degree of node (IDi ) is considered to improve integrated node centrality (INC) for selection of the important nodes, supplementing the deficiencies. Network efficiency is a physical quantity used to describe the ability of the diffusion of information among the networks. Latora and Marchiori (2001) adopted network efficiency to evaluate the resilience of network. Its definition equation is as follows.



E(G) =

i= / j∈G

1 dij

N(N − 1)

(9)

Where E(G) represents the network efficiency of a network model G, and 0 ≤ E(G) ≤ 1; dij denotes the shortest path length between a pair of nodes i, j; N denotes the number of nodes in a network. The load of nodes would affect the network efficiency. It was generally estimated by node’s betweenness, which is shown in Eq. (5) (Motter and Lai, 2002). In a stationary state, the load of each node is smaller than its capacity. When the network is exposed to the removal of a node, the initial load of nodes of network will

X. Li, R. Xiao / Ecological Indicators 74 (2017) 403–413

407

Fig. 3. The constructing flow chart of ID (␨i , Ii , Re, INC).

be changed and lead to the load redistribution over the neighbor nodes. Once the capacity of the node is less than the load of the node, this must induce the node further breaking, triggering a cascade of overload failures and eventually a large drop in the performance of the network, i.e. network efficiency. In a complex network, the load capacity of each node is severely limited by cost. Assuming that the load capacity Ci of node i is proportional to its initial load, it can be expressed as follows. Ci = (1 + ˛)Bv (0)

(10)

Where ␣ is the tolerance parameter, which characterizes the resistance to the attacks and also reflects the construction cost of EIP. It takes the value between 0–1. In the paper, we introduce hypothetical disruptive scenarios on the EIP to assess the change of network efficiency and analyze the resilience of network through removal of critical nodes or companies. 2.4. The complexity of the EIP EIP is designed based on the requirements of clean production, principles of circular economy and industrial ecology, and is composed of the enterprises and the edges denoting the material/energy flow among the enterprises to form industrial symbiosis of shared resources and exchanged by-products. The goal of the EIPs is to seek loop-closing circulation of material, multi-level energy utilization and waste minimization by simulating the natural ecological system and establishing ‘producers-consumersdecomposers’ circulation path in the industrial system (CMEP, 2007). If products, by-products, wastes and materials in the EIP are considered as nodes and the symbiotic exchanges among the products and so on are depicted as edges of EIP. Thus, EIP will form a network, known as ISN, which has the characteristics of complex networks. Generally, the complexity of the network system is mainly manifested in following three aspects (Strogatz, 2001).

(1) As the main organization form in the EIP, the structure of ISN has complex characteristics. Industrial symbiotic relationship refers to the cooperation in the mutual utilization of by-products among different enterprises, that the sum of benefits achieved by working collectively is higher than working as a stand-alone facility (Boix et al., 2012). The exchange of energy, by-products or waste among enterprises in the EIPs forms the vertical and horizontal coupling in the industrial ecological chain. Vertical coupling refers to a relationship formed by hierarchical flow of materials and energy along the ecological chain in order to achieve the multi-level utilization of materials and energy. Horizontal coupling refers to the competition and cooperation relationship among similar enterprises using the same raw materials and products as well as the same main products and by-products. (2) The diversity of products, by-products, wastes and materials types in EIPs makes the complexity of network structure. Because EIPs are made up of a variety of industrial chains, there must be various types of enterprises in EIPs, which yield different products. The enterprises in different industries with different scale act as the “producers”, “consumers” and “decomposers” respectively, in order to establish the ecological cycle in industrial systems. Meanwhile, the products differ in ecological significance in the EIP. (3) The symbiotic relationships in ISN show the complexity influenced by various factors. Owing to various factors, such as changes in national policy, market supply and demand, emergence of new materials and new energy, technical factors, etc., the connected relationships among all enterprises would undergo disruptions, including the connection type, close degree of the symbiotic relationship, presence or absence of a symbiotic relationship, and links direction and weights among the enterprises, etc. Especially, there exist the ecological relationships and non-ecological relationships in the EIPs. Overall, these aspects make the symbiotic network rendering complex

408

X. Li, R. Xiao / Ecological Indicators 74 (2017) 403–413

Fig. 4. The topology of Ningdong CCEIP.

dynamic relationships, and could threaten the stable operation of eco-industrial systems in severe cases. 2.5. Topological structure of symbiosis network in the Ningdong CCEIP In this paper, the network boundary of the Ningdong CCEIP covers all the enterprises involved in the symbiotic transaction of the administrative boundary. Meanwhile, several enterprises outside the CCEIP are added in the network, that exchange resources with the enterprises within CCEIP as supplement components, for an example, coalmine. The products, by-products, wastes and materials are confirmed as nodes. When there is an exchange of materials or energy between any two components, there is an edge between them. On the basis of determined nodes and edges in the CCEIP, the adjacency matrix can be obtained by using the directive dichotomous assessment system to judge whether the interrelationships among 72 nodes exist  or  not in this case. The adjacency matrix is a square matrix X = xij , which is a N × N matrix defined such that



xij =

1 if (i, j) ∈ ε 0 if(i, j) ∈ / ε

(11)

where “1” depicts the existence of an exchange relationship in an adjacent matrix, and “0” depicts the non-existence of any material and energy flow. ε is a set of pairs of different vertices, called edges (Barrat et al., 2008). Obviously, for a directed network the adjacency matrix is not symmetric, xij = / xji ; while for an undirected one to be symmetric, xij = xji . Considering the directional issues in the materials or energy flow among the components, the node “In-degree” and “Out-degree” is used to build Ningdong CCEIP as a directed network. Finally, the paper did not consider the weights of network connections owing to measurement standards of mate-

Table 1 Statistically characteristic parameters of the Ningdong CCEIP. N

M



L

D

C

Lrand

Crand

72

224

6.2222

3.5660

9.0000

0.1542

2.3394

0.0864

rials and energy exchange could not be unified. The NetDraw tool in Ucinet6.0 software is used to get the topology diagram of Ningdong CCEIP network, which is shown as Fig. 4. The nodes, which indicate the products, by-products, and waste of Ningdong CCEIP, are numbered. 3. Results and analysis 3.1. Characteristic analysis of topology in complex network Analyzing the topological properties of network is contributed to further study the role of the key nodes during EIP’s development and operation. The topological parameters of network are used to study network features, mainly including degree, degree distribution, average shortest path length, and clustering coefficient, etc. (Watts, 1999). MATLAB7.0 software is adopted to calculate the above-mentioned statistical parameters, and the results are shown in Table 1. From Table 1, the number (N) of the nodes in the EIP is 72. The edges (M) are 224. The average degree is 6.2222. The average shortest path length (L) and the diameter (D) are 3.566 and 9.0 respectively. The basic topological features, i.e. scale-free and small-world one, are analyzed in the following parts. 3.1.1. Validation of the scale-free characteristics Barabasi and Albert (1999) defined scale-free network in which the node degree distribution satisfied the power-law distribution. Therefore, an important feature of the scale-free network is that

X. Li, R. Xiao / Ecological Indicators 74 (2017) 403–413

409

This phenomenon is also known as the phrase: ‘the rich get richer’ (Barabasi and Albert, 1999). 3.1.2. Validation of small world characteristics Based on the analysis of the changes in the relationships among the clustering coefficient, the average shortest path length and the reconnect probability P in a WS small world network, it can be found that the small world network model is featured in a shorter average path length and a much higher clustering coefficient. Sporns et al. pointed out that when the network met Eq. (11), the network was featured in the small world (Sporns et al., 2007). C L > Crand Lrand

Fig. 5. (a) Out-degree distribution in Ningdong CCEIP. (b) In-degree distribution in Ningdong CCEIP.

the node degree satisfies the power-law distribution, i.e. p(k)∼k− . That means the hub nodes appear in the network. In this paper, the degree distribution is investigated in the condition of double logarithmic coordinates in order to validate the scale-free characteristics of the industrial network. If there is a linear relationship between logp(k) and logk satisfying log p(k) = − log k + a, it can be proven that the network satisfies the condition of the powerlaw distribution, i.e., p(k)∼k− . The process can be implemented by MATLAB7.0 software. Fig. 5(a) and (b) shows the degree distribution of the network of Ningdong CCEIP in the condition of the double logarithmic coordinates. Fig. 5(a) illustrates the out-degree distribution of the network, where the regression coefficient  is 1.226. Fig. 5(b) indicates the in-degree of the network, where the regression coefficient  is 1.4510. The regression coefficients are a bit less than the measuring results of a large number of empirical studies. The results clearly show that there are linear relationships between logp(k) and logk of in-degree and out-degree distribution in the network. That implies, the out-degree and in-degree distribution of network in Ningdong CCEIP is in line with power-law distribution, and the network has a scale-free feature. This indicates that the network has a small amount of hub nodes (i.e. a greater degree of nodes), which take a dominant position in the network, affecting a relatively small degree of nodes in the EIP. At the same time, because the scale-free network is featured in ‘preferential attachment’, new enterprises in the EIP are more inclined to link with anchor ones.

(12)

Where Crand is the clustering coefficient of random network (ER), ; Lrand is the average path length of random network Crand = N ln N ; N is the number of nodes in random network, (ER), Lrand = ln and here it’s 72; < k > is the average degree of random network. In this paper, Ucinet6.0 software has been adopted to build a random network (ER), which is endowed with the same node number and the same relationships to Ningdong CCEIP. If the average path length of real network is close to that of random network, and the clustering coefficient of real network is significantly greater than the one of random network, it can be proven that Ningdong CCEIP is featured in small world (Watts and Strogatz, 1998). The calculated results of Crand and Lrand are shown in Table 1.The results show that the average shortest path length of the Ningdong CCEIP is approximately same as the corresponding random network, while the clustering coefficient is greater than the one of random network, which satisfies Eq. (12). So it can be said that Ningdong CCEIP does have the characteristics of small world. This indicates that the enterprises in the network can conveniently exchange the materials/energy. The larger the value of the clustering coefficient, the higher the degree of conglomeration. Therefore, the fostering and development of the enterprise groups with a higher clustering coefficient can be studied, which are deemed as core industries or advantageous ones. Moreover, Kühnert et al. (2006) proposed that the average path length represented the delivery time of products in the network. This is to say, the smaller average path length, the higher efficiency of exchange of materials or energy among the enterprises. 3.2. Network analysis based on node-level metrics Fig. 6 presents the node-level metrics of the five nodes for Ningdong CCEIP. The results in Fig. 6 illustrate the importance of the five products in the network as indicated by several network metrics. There is an apparent trend that waste residue has the highest total degree, in-degree, node betweenness and integrated node centrality (INC) suggesting its high importance in the system. Waste residue, exhaust gas and wastewater have high in-degree because they come from most of industries, and relatively low outdegree because direct several nodes in the network. And they have high total degree indicating importance in the network as conveyed by degree centralities. PVC has a high out-degree because it is responsible for supply of raw materials to other industries. At the same time, steam has a high out-degree because it directs to other nodes in the network. However, PVC and thermal power have a much comparatively much lower importance in the network for total degree. Waste residue and exhaust gas also show as the most centrally located node in the work based on node betweenness, indicating the important nature of the industry, as a point of vulnerability for the network. It is worthwhile mentioned that waste residue and exhaust gas have been obtained well exchange between the industries reflected in their high betweenness, including PVC. However,

410

X. Li, R. Xiao / Ecological Indicators 74 (2017) 403–413

total degree in-degree

out-degree betweenness

INC

1.0

network metrics (normalized to 1)

0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1

m st ea

as

te w

at

PV C

er

as us tg w

w

as

te

ex ha

re

s id

ue

0.0

Fig. 6. Network analysis based node-level metrics for Ningdong CCEIP.

important degree of node (normalized to 1)

1.0 0.9 0.8

a certain extent reflects the importance of a node in the topological structure of network, but it cannot accurately indicate the importance of the node in the ecological symbiosis network. Compared with waste residue and wastewater, effective utilization and treatment of exhaust gas are more crucial to green production, safety operation and environmental protection. Waste residue treatment mainly includes carbide slag, coal ash, cinders and so on, as the raw materials, but has a limited role in wastes recycling in Ningdong CCEIP. To meet the environmental requirement of EIPs enacted by the Chinese government, the wastewater produced by all enterprises must be discharged through pipelines and treated in the wastewater treatment plant, which makes the wastewater treatment plant bound up with all enterprises. However, recycling treatment water is very limited, which lead to relatively low ecological factor and out-degree. PVC is at the central position in the salt chemical industry chain, which has some downstream firms linked with. Compared with wastewater, the higher importance of PVC shows PVC play critical role in the production process in the EIP, and eco-efficiency of wastewater need to further be improved. Steam from thermal power is fully utilized. However, other plants producing POM, PET, PTT are not listed as top based on the ID. Actually, POM is the main product in Ningdong CCEIP. It is noted that further improving ecological efficiency in the industrial chain is taken into account. Therefore, the important degree (ID) of nodes can not only inform us the important node in the present EIP, but also hint the improvement direction for eco-efficiency in the future.

0.7 0.6

3.4. Resilience analysis of Ningdong CCEIP

0.5 0.4 0.3 0.2 0.1

m ste a

as te w at er w

PV C

ue re sid w as te

ex ha us tg

as

0.0

Fig. 7. Important degree of nodes based ecological effect and INC.

there is inconsistence in sequencing the five nodes for total degree and betweenness, mainly wastewater and PVC. Therefore, INC is needed to introduce to analyze the importance of the nodes in the network. Considering the two factors based node-level metrics, i.e. node degree and beweenness, waste residue and exhaust gas still the most important nodes in the network according to the values of INC. From Fig. 6, PVC, wastewater and steam successively rank left behind. From network analysis based node-level metrics, it can be seen that waste residue and exhaust gas are the vulnerable nodes for developing resilience networks. However, whether wastewater rank behind PVC needs to further be verified. 3.3. Importance of nodes based ecological factor and INC Fig. 7 presents the results for the important degree (ID) of nodes on the basis of ecological factor and integrated node centrality (INC). According to Fig. 7, exhaust gas, waste residue and PVC rank top three, followed by wastewater and steam. Compared to the sequence in accordance with the value of INC, the sequence in accordance with ID is more consistent with the actual situation of Ningdong CCEIP. That is to say, the value of the INC to

The complete removal of the more important nodes as a result of targeted or untargeted disruption has a higher impact on the structure of the network than the removal of less important ones (Chopra and Khanna, 2014). Fig. 8 presents the changing trends of network efficiency for Ningdong CCEIP under sequential removal of important nodes based ID, without taking into account disruption propagations. It is indicated that a removal of only about 10% of nodes in Ningdong CCEIP will lead to the drop in the network efficiency sharply. That is to say, the resilience of the network will decrease, and corresponding vulnerability of the network will increase. And only the removal of top 10% node contributes to 60% decrease of network efficiency in the studied case. Therefore, Ningdong CCEIP is preferentially attached and has more richly connected, anchor firms. The result reflects the general findings that in the complex network, the heterogeneous network are relatively vulnerable to targeted removals of most important nodes (Albert et al., 2000). However, as should be expected Ningdong CCEIP are less resilient when facing disruptions. For example, when the exhaust gas treatment shuts down due to malfunction, there are great problems in the operation of EIP, not only failing to satisfy environmental protection index of EIP, but also endangering the safety of the surrounding residents. Fig. 9 shows a plot of the network efficiency versus tolerance parameter when the most important node (exhaust gas) is removed. We observe that when a is equal to 0, network efficiency declines to the minimum value, i.e., 0.012. With the increase of a (0–1), network efficiency keeps rising. When a is equal to 1, network efficiency rises to 0.042. It indicates that the network has the vulnerability for targeted disruptions. When the most important node is suffered from removal, the connectivity of the whole network will be affected greatly. In the actual situation, due to the restriction of the cost, a usually is less than 0.3. When a is equal to 0.3, network efficiency is 0.014. The research has shown that protecting the most important nodes is critical to safeguard the potential “vulnerability” in the development of EIPs.

X. Li, R. Xiao / Ecological Indicators 74 (2017) 403–413

411

Fig. 8. The changing trend of network efficiency under disruptions.

Fig. 9. Network efficiency versus tolerance parameter.

4. Discussion 4.1. The potential role of network theory for understanding ISN Since most of the industrial symbiotic networks may be a result of social interactions between managers and owners of industries, the EIPs may not be strategically planned and be spontaneous in nature (Chopra and Khanna, 2014). It is necessary to shed light on the structural characteristics of ISN, promoting to systematically design and develop the networks. Industrial ecosystems have been considered as complex adaptive self-organizing systems. Complex

network theory can provide a comprehensive analysis framework to understand the patterns of organization in the industrial ecosystems. Our topological structure analysis of EIP not only is based on regular statistical parameters, but also introduces a new indicator, i.e. important degree of nodes. The indicator integrates topological structure and ecological feature, differing from other studies (Chopra and Khanna, 2014; Domenecha and Daviesa, 2011; Zeng et al., 2013), which describe the important nodes using node degree, node betweenness, or ecological effect and so on. Our work contributes to promote the application of network theory in the ISN.

412

X. Li, R. Xiao / Ecological Indicators 74 (2017) 403–413

Meanwhile, more suitable measure to evaluate the node importance put forward can help us better understanding the topological and ecological feature of EIP, and making an effort to operate and manage the EIP in order to achieve the comprehensive utilization of materials, energy and information among enterprises. 4.2. The importance of network resilience for EIP During the operation process of EIP, the unanticipated perturbations may affect even one enterprise (or production facilities), and lead to a domino effect, resulting in cascading impacts on the rest of the network (Boons and Spekkink, 2012), which makes the system display great vulnerability. Until now, the potential “vulnerability” of the system is the main problem affecting the function of the entire chain in the development of EIPs (Tudor et al., 2007). To recommend the design strategies of EIPs, it is very necessary to evaluate and improve resilience of the EIP. Network resilience is mainly related to topological structure, namely the homogeneous and heterogeneous network. It is worthwhile noted that the removal of top 10% node contributes to 60% decrease of network efficiency in the studied case. And the most important node removed leads to greatly effect on the network though concerning tolerance parameter. The analysis in the paper provides some understanding the strategies for designing a resilient EIP. Because resilience is the ability of a system to respond to change, comprehensively analyzing the possible perturbation process is crucial for developing adaptive capacity in an EIP from topological structure and ecological feature. To track the evolution of resilience in an EIP, not only snapshot analysis, but also time trend needs to be concentrated on in order to develop novel mechanism to avoid disruptions, improve the resilience of EIP and safeguard the stable operation. 5. Conclusions In this paper, complex network theory is adopted to analyze network topological characteristics from the perspective of resilience in order to advance the sustainable development of EIP. Firstly, the statistical parameters are applied to analyze the topological properties of symbiotic network. Secondly, we propose a novel model to select the important nodes in EIPs. Finally, we analyze the changing trend of resilience for Ningdong CCEIP under sequential removal of important nodes based on network efficiency. Based on the current studies, the following conclusions are drawn: in the case of Ningdong CCEIP in China, as a directed graph, based on the values of statistical parameters, the studied network is verified with the scale-free characteristics and small world ones. Compared with the node-level metrics, the important degree of node considering ecological factor can better reflect the importance of the node in the overall network. The removal of top 10% node contributes to 60% decrease of network resilience. In this paper, we analyze the network characteristics only from the topological structure not considering the concrete material flows as weighs for the nodes, in part to due to a lack of data for the volume of materials but also to simplify our exploration. Actually, the analysis of the resilience based on the kinds of flows (including material, energy and information flow) of the real network should better represent the real relationships among the nodes and be closed to the actual situation of EIP (Xiao et al., 2016), because the kinds of flows directly affect the economic and ecological benefits. In the future, emergy analysis will be introduced to unify the kinds of flows into the same form of energy, usually solar energy (Odum, 1996). And we will strive to premeditate the emergy flows as weights for the nodes and explore how different nodes affect the resilience information in the weighted-network. Mean-

while, because nodes with more network links do not represent their importance in the network (Hu et al., 2015), comprehensively addressing more types of node attributes such as ecological effect, economic outcomes will contribute to explore the structure regularities embedded in the EIP. The further research will furtherly reveal the reasons of protecting the most importance nodes in the stable operation of EIP. Acknowledgement We thank Editor Felix Müller and the anonymous reviewers for their pertinent comments that greatly improve the original manuscript. This research was supported by the National Natural Science Foundation of China (41501183, 71171089). References Albert, R., Barabási, A.L., 2002. Statistical mechanics of complex networks. Rev. Mod. Phys. 74 (1), 47–97. Albert, R., Jeong, H., Barabási, A.L., 2000. Error and attack tolerance of complex networks. Nature 406, 378–382. Barabasi, A.L., Albert, R., 1999. Emergence of scaling in random networks. Science 286, 509–512. Barrat, A., Barthelemy, M., Vespignani, A., 2008. Dynamical Processes on Complex Networks. Cambridge University Press, New York. Behera, S.K., Kim, J.H., Lee, S.Y., Suh, S., Park, H.S., 2012. Evolution of ‘designed’ industrial symbiosis networks in the Ulsan Eco-industrial Park: ‘research and development into business’ as the enabling framework. J. Clean. Prod. 29–30, 103–112. Boix, M., Montastruc, L., Pibouleau, L., Azzaro-Pantel, C., Domenech, S., 2012. Industrial water management by multiobjective optimization: from individual to collective solution through eco-industrial parks. J. Clean. Prod. 22 (1), 85–97. Boons, F., Spekkink, W., Mouzakitis, Y., 2011. The dynamics of industrial symbiosis: a proposal for a conceptual framework based upon a comprehensive literature review. J. Clean. Prod. 19, 905–911. Côté, R.P., Cohen-Rosenthal, E., 1998. Designing eco-industrial parks: a synthesis of some experiences. J. Clean. Prod. 6 (3–4), 181–188. Chen, Y., Wang, X.L., Bu, J.Z., Tang, B.Z., Xiang, X., 2016. Network structure exploration in networks with node attributes. Physica A 449, 240–253. Chertow, M., Lombardi, D., 2005. Quantifying economic and environmental benefits of co-located firms. Environ. Sci. Technol. 39 (17), 6535–6541. Chertow, M.R., 2000. Industrial symbiosis: literature and taxonomy. Annu. Rev. Energy Environ. 25, 313–337. Chinese Ministry of Environmental Protection (CMEP), 2007. Guide for Established of Eco-industrial Parks Planning HJ/T 409-2007. China Standard Press, Beijing, China (In Chinese). Chopra, S.S., Khanna, V., 2014. Understanding resilience in industrial symbiosis networks: insights from network analysis. J. Environ. Manage. 141, 86–94. Domenecha, T., Daviesa, M., 2011. Structure and morphology of industrial symbiosis networks: the case of Kalundborg. Procedia Soc. Behav. Sci. 10, 79–89. Frosch, R.A., Gallopoulos, N.E., 1989. Strategies for manufacturing. Sci. Am. 261 (3), 144–152. Gao, J.X., Barzel, B., Barabási, A.L., 2016. Universal resilience patterns in complex networks. Nature 530, 307–312. Gibbs, D., Krueger, R., 2005. Exploring local capacities for sustainable development. Geoforum 36 (4), 407–409. Hong, L., Ouyang, M., Peeta, S., et al., 2015. Vulnerability assessment and mitigation for the Chinese railway system under floods. Reliab. Eng. Syst. Saf. 137, 58–68. Hu, P., Fan, W.L., Mei, S.W., 2015. Identifying node importance in complex networks. Physica A 429, 169–176. Kühnert, C., Helbing, D., West, G.B., 2006. Scaling laws in urban supply networks. Physica A 363 (1), 96–103. Korhonen, J., Snä kin, J.P., 2005. Analysing the evolution of industrial ecosystems: concepts and application. Ecol. Econ. 52 (2), 169–186. Korhonen, J., Seager, T.P., 2008. Beyond eco-efficiency: a resilience perspective. Bus. Strategy Environ. 17, 411–419. Korhonen, J., Wihersaari, M., Savolainen, I., 1999. Industrial ecology of a regional energy supply system: the case of Jyvaskyla region, Finland. Greener Manage. Int. 26, 57–67. Latora, S., Marchiori, A., 2001. Attack vulnerability of scale-free networks due to cascading breakdown. Phys. Rev. E 70 (3), 5101–5114. Liu, Z.H., Jiang, C., Wang, J.Y., Yu, H., 2015. The node importance in actual complex networks based on a multi-attribute ranking method. Knowledge-Based Syst. 84, 56–66. Liwarska-Bizukojc, E., Bizukojc, M., Marcinkowski, A., Doniec, A., 2009. The conceptual model of an eco-industrial park based upon ecological relationships. J. Clean. Prod. 17, 732–741. Lowe, E.A., Koenig, A.W., 2006. Eco-Industrial Training Manual for KNCPC EIP Pilot Projects – Train the Trainers Workshop. Indigo Development, Oakland, CA.

X. Li, R. Xiao / Ecological Indicators 74 (2017) 403–413 Motter, A.E., Lai, Y.C., 2002. Cascade-based attacks on complex networks. Phys. Rev. E 66 (2), 065102. Odum, H.T., 1996. Environmental Accounting: Emergy and Environmental Decision Making. John Wiley and Sons, New York. Park, Behera, 2014. Methodological aspects of applying eco-efficiency indicators to industrial symbiosis networks. J. Clean. Prod. 64, 478–485. Patel, R., Modi, B., Patwari, S., Gopichandran, R., Wilderer, M., 2001. Aspect of e eco-industrial networking exercise at the Naroda industrial estate. In: Proceedings of International Conference on Industrial Park Management: New Strategies for Industrial Development, April 3–6. Manila, Philippines. Sporns, O., Honey, C.J., Kötter, R., 2007. Identification and classification of hubs in brain networks. PLoS One 2 (10), e1049, http://dx.doi.org/10.1371/journal. pone.0001049. Strogatz, S.H., 2001. Exploring complex networks. Nature 410, 268–276. Tammemagi, H., 1999. The Waste Crisis, Landfills, Incinerators and the Search for a Sustainable Future. Oxford University Press, Oxford, UK. Tan, R.R., Col-long, K.J., Foo, D.C.Y., Hul, S., Ng, D.K.S., 2008. A methodology for the design of efficient resource conservation networks using adaptive swarm intelligence. J. Clean. Prod. 16 (7), 822–832. Tudor, T., Adam, E., Bates, M., 2007. Drivers and limitations for the successful development and functioning of EIPs (eco-industrial parks): a literature review. Ecol. Econ. 61 (2–3), 199–207. United Nations Environmental Programme (UNEP), 2000. Global Environment Outlook 2000. Earthscan, London. Van Beers, D., Corder, G., Bossilkov, A., van Berkel, R., 2007. Industrial symbiosis in the Australian minerals industry: the cases of Kwinana and Gladstone. J. Ind. Ecol. 11 (1), 55–72. Venta, G.J., Nisbet, M., 1997. Opportunities for Industrial Ecological Parks in Canada. Case Study of Sarnia–Lambton Industrial Complex. Environment Canada, Ottawa.

413

Watts, D.J., Strogatz, S.H., 1998. Collective dynamics of small-world networks. Nature 393, 440–442. Watts, D.J., 1999. Networks, dynamics, and the small world phenomenon. Am. J. Sociol. 105, 493–592. Wright, R.A., Côté, R.P., Duffy, J., Brazner, J., 2009. Diversity and connectance in an industrial context the case of Burnside industrial park. J. Ind. Ecol. 13 (4), 551–564. Xiao, Z.D., Zhou, G.H., Wang, B., 2012. Using modified Baraba´ısi and Albert model to study the complex logistic network in eco-industrial systems. Int. J. Prod. Econ. 140 (1), 295–304. Xiao, Z.D., Cao, B.B., Sun, J., An Zhou, G.H., 2016. Culture of the stability in an eco-industrial system centeredon complex network theory. J. Clean. Prod. 113, 730–742. Xu, G.Z., 2000. Systems Science. Shanghai Science and Technology Education Press, Shanghai, China (In Chinese). Zeng, Y., Xiao, R.B., Li, X.M., 2013. A resilience approach to symbiosis networks of ecoindustrial parks based on cascading failure model. Math. Prob. Eng., 1–11. Zhang, Y., Zheng, H.M., Fath, B.D., 2015. Ecological network analysis of an industrial symbiosis system: a case study of the Shandong Lubeieco-industrial park. Ecol. Modell. 306, 174–184. Zhu, J.M., Ruth, M., 2013. Exploring the resilience of industrial ecosystems. J. Environ. Manage. 122, 65–75. Zhu, Q., Lowe, E., Wei, Y., Barnes, D., 2007. Industrial symbiosis in China: a case study of the Guitang group. J. Ind. Ecol. 11 (1), 31–42. Zhu, L., Zhou, J.R., Cui, Z.J., Liu, L., 2010. A method for controlling enterprises access to an eco-industrial park. Sci. Total Environ. 408, 4817–4825.