Ecological Indicators 51 (2015) 53–58

Contents lists available at ScienceDirect

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

Dynamic assessment of environmental damage based on the optimal

clustering criterion – Taking oil spill damage to marine ecological
environment as an example
Tong Yang *
School of Finance & Public Administration, Anhui University of Finance and Economics, Bengbu 233030, China

A R T I C L E I N F O A B S T R A C T

Article history: Aim: ng at the evaluation of environmental damages, we proposed a dynamic evaluation approach based
Received 28 April 2014 on the optimal cluster criterion. Firstly, a method for sample data standardization was introduced. After
Received in revised form 15 September 2014 determining the measurement of damage grade, we applied the system cluster analysis approach to
Accepted 20 September 2014
classify the grades of corresponding environmental pollution events. Then, the optimal cluster level was
evaluated based on the optimal clustering criterion. By using the marine environmental damages caused
Keywords: by 17 marine oil spill events as a sample, we tested the dynamic evaluation method proposed in this
Environmental damage
article with its practicability. Further, by comparing it with some traditional damage evaluation methods,
Optimal cluster
Dynamic evaluation
we found that their evaluation results were consistent. All these have shown that the dynamic evaluation
Marine oil spill approach proposed in this paper can meet the requirement of environmental damage evaluation. Finally,
Consistency check directions of future researches were pointed out.
ã 2014 Elsevier Ltd. All rights reserved.

1. Introduction In current studies, the scholars introduced different methods to

Along with the entry of social economy into fast industrialization
stage, contradictions between human beings and the environment
gradually appeared. Environment pollution incidents happened
from time to time (Peng et al., 2013), which would definitely make
damages to the environment. Environmental damage refers to
various kinds of disasters happening during or after environmental
pollution events, including direct damages to regional ecological
environment and natural resource and indirect damages of expenses
occurred for taking necessary and proper measures to prevent
expansion of the pollutions, repairing or recovering damaged
ecological environment, and so on (Gastineau and Taugourdeau,
2013; Martin-Ortega et al., 2011). Damage evaluation is the first
question to be faced after the occurrence of environmental damages.
It is the basis for a series of issues such as compensation and pollution
treatment afterwards (Burger, 2008). Therefore, we need to find out
an effective method for environmental damage evaluation which can
be used to dynamically grade environmental damages in different
environmental areas and for corresponding environmental pollution
events. Hence, dynamic evaluation of environmental damage has
become a hot research subject at present.
* Tel.: +86 13855200689.
E-mail address: ivytong520@163.com (T. Yang).
make environmental damage evaluations. For example, Hanson et al.
(2013) introduced the Habitat Equivalence Analysis (HEA) into the
environmental damage evaluation. They proposed a compensation
quantity model, a damage model and a compensation scale model.
These models can be used to simulate marine oil spill accidents,
analyze the evaluation scope and the damage ways. Although HEA is
a good tool to measure the materialized loss and evaluate ecological
value loss, it cannot rank the environmental damage events and
decide damage grades (Ma and Qiang, 1988). Vaissiere et al. (2013)
introduced the equivalence analytical method into the environmen-
tal damage evaluation and carried out a case study to the
environmental damages caused by unexpected water pollutions
to a river. They tried to evaluate environmental damage through
constructing simulated equivalent ecology, but rationality of this
method still needs to be further proved (Rodríguez-Labajos and
Martínez-Alier, 2013). Gilbuena et al. (2013) adopted the utility
function method and the shadow projection approach when
evaluating environmental damage. However, utility function meth-
od is likely to be affected by subjective factors and shadow projection
approach can be greatly affected by the values of damaged marine
areas and water depth when evaluating environmental damages
caused by oil spill accidents (Fallah-Mehdipour et al., 2011). Brody
et al. (2012) carried out a preliminary evaluation to the ecological
environmental damages resulted from coal mining in Shanxi
province by using the economic loss coefficient of ecological

http://dx.doi.org/10.1016/j.ecolind.2014.09.033
1470-160X/ ã 2014 Elsevier Ltd. All rights reserved.
54 T. Yang / Ecological Indicators 51 (2015) 53–58

environment and preliminarily predicted the possible damages handling them, because it's meaningless to compare data when the
resulted from coal mining to the environment in the future. However, dimensions of index vectors are different or quantity levels vary
the method mentioned in this literature was only used to evaluate greatly. For convenience, we assume that all acquired samples
the economic losses resulted from environmental damages. The were expressed by matrix X.
marine ecological service function evaluation model put forward by 2 3
x11 :::x1j :::x1m
Kennedy and Cheong (2013) is weakly operable because value and 6 ::::::::::::::: 7
time of ecological function loss are hard to decide. Also, there are 6 7
X¼6 7
6 xi1 :::xij :::xim 7 (1)
some statistical methods can be used to make environmental 4 ::::::::::::::: 5
damage evaluation. For example, matter-element analysis, the xn1 :::xnj :::xnm
entropy method and the main component method. All these
The matrix means that the sample size is n and each sample
statistical method can rank the environmental pollution events
contains m damage index. xij refers to the value of the jth
effectively, but they all cannot dynamically figure the disaster levels
damage index of the ith sample. Then, to each kind of index
of the environmental pollution events. Hence, high effectiveness and
with the same type, its sample variance will be as shown in
accuracy of evaluation cannot be guaranteed.
Formula (2):
Although there are many researches on evaluation of environ-
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
mental damage, but there is still not a systematic method which can u
u 1 X n
comprehensively evaluate the environmental damage in a dynamic Sj ¼ t ðx  xj Þ2 (2)
manner. To solve this problem, this article is written with efforts to n  1 i¼1 ij
dynamically classify the grade of environmental damage and identify Xn
xj ¼ 1n xij refers to the mean value of index value of the jth type.
the degree of damage to the to-be-evaluated environmental i¼1
pollution event under corresponding conditions. The optimal cluster
Based on the above assumption, following formula is put
approach (Mulder, 2013; Li et al., 2013) is introduced into the
forward to standardize the acquired sample data:
environmental damage evaluation in this article. By incorporating
8
this approach, problems about dynamic evaluation of environmental < xij  xj
 Sj 6¼ 0
damage can be effectively solved. Further, some other statistical xij ¼ Sj (3)
:
methods are also incorporated into our evaluating approach, the 0 Sj ¼ 0
matter-element analysis method corporate with the proposed Through the above changes, mean value of data is 0 and the
cluster analysis are used to make comprehensive analysis of the standard deviation is 1. Thus, the effect of dimension can be
environmental pollution events, specifically they can be used to eliminated. Even though samples are changed, results can stay
identify the disaster grade of the environmental pollution events. stable.
The entropy method and the main component method are
incorporated to rank the damages of the environmental pollution
events. Additionally, marine ecological environmental damages 3. Measurement of damage grades based on optimal cluster
caused by 17 oil spill accidents is set up to test this environmental
dynamic evaluation method. Results of optimal cluster approach, We will put forward the measurement of damage degree at first
matter-element analytical method, entropy evaluation method and and then propose the corresponding cluster analysis method to
principal component evaluation method are compared to check the classify environmental damages caused by each accident and
consistency of the proposed comprehensive environmental damage decide the number of categories of marine ecological environ-
evaluation method. mental damages caused by oil spill accidents.

2. Data standardization 3.1. Measurement of damage grade and system cluster analysis

We need to standardize the sample data of marine ecological Before carrying out system cluster method analysis, we need to
environmental damages resulted from oil spill accidents before decide the measurement of environmental damage grade. The

Table 1
Indexes for marine ecological environmental damages caused by oil spills.

Event Economic loss (unit: RMB 100 million yuan) Fishery disaster (unit: 10,000 ha) Pollution-affected population (unit:
number 10,000 persons)
Direct economic Fishery industry economic Polluted ocean Polluted fishery industry
loss loss area area
1 237.5314 72.7394 216.1040 77.2733 2047.3200
2 65.9884 83.1057 52.5333 10.4000 220.0000
3 184.8036 83.3782 217.4188 48.4222 1464.9000
4 214.3930 65.3519 105.0759 39.574 1148.0000
5 85.9619 100.1141 68.8140 12.4867 464.8400
6 101.1890 94.5755 16.5000 3.0000 196.0000
7 86.6274 87.2703 163.9000 24.8000 1377.0000
8 229.5515 78.7274 286.3800 89.9100 2393.6700
9 51.1839 89.4617 23.7800 0.0000 387.8700
10 158.7388 76.0180 153.5000 40.5700 1640.1000
11 51.4421 98.0325 59.2480 15.1140 640.1100
12 116.2253 72.9618 161.6700 55.4500 1881.9200
13 160.8527 65.8988 67.8100 11.1400 814.6000
14 110.0926 99.4163 38.2600 4.3700 644.3400
15 83.7927 87.8358 108.4000 13.4000 1288.7000
16 64.8935 107.1756 6.1200 3.0400 538.7000
17 32.8705 122.9369 32.3500 0.0000 398.5500
 T. Yang / Ecological Indicators 51 (2015) 53–58 55

Table 2
Results of standardization of data.

Event number Economic loss Fishery disaster Pollution-affected population

Direct economic loss Fishery industry economic loss Polluted ocean area Polluted fishery industry area
1 1.7916 0.9558 1.3549 1.8361 1.4874
2 0.8183 0.2778 0.6323 0.5779 1.1899
3 0.9894 0.2600 1.3709 0.7946 0.6340
4 1.4396 1.4390 0.0060 0.4752 0.1697
5 0.5144 0.8347 0.4345 0.5026 0.8312
6 0.2827 0.4724 1.0701 -0.8450 1.2251
7 0.5043 0.0054 0.7207 0.0581 0.5052
8 1.6702 0.5642 2.2087 2.2923 1.9948
9 1.0435 0.1379 0.9817 0.9533 0.9440
10 0.5928 0.7414 0.5943 0.5112 0.8907
11 1.0396 0.6985 0.5507 0.4077 0.5744
12 0.0540 0.9413 0.6936 1.0483 1.2450
13 0.6250 1.4032 0.4467 0.5512 0.3187
14 0.1473 0.7890 0.8057 0.7956 0.5682
15 0.5474 0.0316 0.0464 0.4696 0.3759
16 0.8349 1.2965 1.1962 0.8436 0.7230
17 1.3221 2.3274 0.8775 0.9533 0.9283

measurement can be divided into sample distance measurement In the formula, np and nq respectively refers to the quantities of
and between-class distance measurement. samples in types p and q; xp and xq refers to the mean value of each
The sample distance measurement adopted in this article is the index of each sample in types p and q respectively.
included angle cosine distance as shown in Formula (4): After the selection of measurement, corresponding cluster
Pm 2
analytical method can be used to classify the selected samples.
j¼1 ðxij xi0 j Þ Basic thoughts in this method are as follows: let n samples to form
consineðxi ; xi Þ ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Pm 2 Pm 2 (4)
ð j¼1 xij Þð j¼1 xi0 j Þ a type spontaneously and then measure the similarity at which
time the between-class distance and the between-sample distance
are equivalent; combine two samples with the least measurement;
consineðxi ; xi Þ 2 ½0; 1 refers to the similarity of the ith sample to the
calculate the between-class distance based on the acquired
i0 th sample. If consineðxi ; xi Þ ¼ 1, it means that those two samples
distance measurements and then combine types as per the
are extremely similar. If the value is 0, it means that those two
minimum distance standard until there is only one type of sample.
samples are irrelevant.
Process of cluster can be expressed by the cluster hierarchical
The between-class distance measurement adopted in this
diagram.
article is the distance of sum of deviation squares as shown in
Formula (5):
3.2. Hierarchical optimal cluster criterion function
np nq
D2pq ¼ ðxp  xq ÞT ðxp  xq Þ (5)
np þ nq After deciding the measurement of damage grade and
classification and clustering of samples, we need to decide the
optimal damage grade for the marine environmental damage
[(Fig._1)TD$IG]

Fig. 1. Cluster dendrogram of marine ecological environmental damages caused by oil spill accidents.
 56 T. Yang / Ecological Indicators 51 (2015) 53–58
[(Fig._2)TD$IG]

Fig. 2. Optimal cluster number.

caused by oil spill accidents. Here, the optimal cluster criterion direct economic loss (unit: RMB 100 million yuan), the fishery
function is introduced as shown in Formula (6): industry economic loss (unit: RMB 100 million yuan), the polluted
X marine area (unit: 10,000 ha), the polluted fishery industry area
PG ¼ ðxi  xk ÞT ðxi  xk Þ (6) (unit: 10,000 ha) and the pollution-affected population (unit:
i2Gk 10,000 persons) are used. Please refer to the Table 1 for example.
This formula refers to the sum of squares of between-type We will carry out the dynamic evaluation to the marine ecological
deviations within samples in the Gk type of which xk is the center. If environmental damages caused by oil spills through the collected
the sum of squares of between-type deviations is small, it means sample data and by using the method mentioned in Section 3.
that samples in this type are similar. However, the number of Results of standardization of sample data by using the method
categories c will then become very large. Hence, following function mentioned in Section 2 are stated in Table 2.
will be considered based on the consideration of the sum of Results of cluster analysis of samples by using the measurement
between-type deviation squares: and the system cluster analytic method mentioned in Section 3.1
 3  3 X are summarized in a cluster diagram, Fig. 1.
c c From the figure we can see clearly each process of merging and
f ðcÞ ¼ PG ¼ ðx  xk ÞT ðxi  xk Þ (7)
cþ2 c þ 2 i2G i sorting by using the system cluster analytic method and the
k
number of categories of clusters in each stage. The scatter diagram
Set:
of (c, f(c)) can be made as per the optimal cluster criterion function
 3 in Formula (7). Please refer to Fig. 2 for details.
c
gðcÞ ¼ (8) Judging from the analytic results shown in the figure, the
cþ2
optimal cluster number is 4. Therefore, marine ecological
Because g0 (c) > 0 when c  1, g(c), is the monotone increasing environmental damages caused by oil spill accidents can be best
function of c while PG is the monotone decreasing function. Hence, divided into 4 grades, respectively huge damage, major damage,
there will be a maximum value of f(c) = g(c)PG which is the optimal moderate damage and small damage.
cluster number needed. Grade of marine ecological environmental damage caused by
When solving practical marine environmental damages each oil spill accident can be worked out as per the optimal
resulted from oil spill accidents, different c values will turn out grade number and the cluster dendrogram. Results are shown in
different f(c) values. Then from the scatter diagram of (c,f(c)) we Table 3.
can see that the number of categories corresponding to max f(c) is From the table we can know that the nos. 1, 3, 4, 8 and 10
just the optimal cluster number. oil spill accidents have caused huge damages to the marine

4. Empirical analysis
Table 3
4.1. Dynamic evaluation of environmental damage based on the Cluster of grade of marine ecological environmental damage caused by each oil spill
accident.
optimal cluster
Grade Serial number of oil spill accident
According to the features of environmental damages and the Grade 1 (huge damage) 1, 3, 4, 8, 10
availability of statistical index data, data about marine ecological Grade 2 (major damage) 5, 6, 7
environmental damages caused by 17 oil spill accidents in recent Grade 3 (moderate damage) 2, 9, 11, 14, 15, 16, 17
Grade 4 (small damage) 12, 13
five years have been adopted in this article. Five indexes, the
 T. Yang / Ecological Indicators 51 (2015) 53–58 57

Table 4
Comprehensive analysis results of marine ecological environmental damages caused by oil spill accidents.

Accident no. Optimal cluster Matter-element evaluation Entropy method Main component method
1 1 2 4 3
2 3 3 12 11
3 1 2 6 6
4 1 2 5 2
5 2 2 2 5
6 2 2 3 4
7 2 2 8 8
8 1 1 1 1
9 3 4 15 16
10 1 2 7 7
11 3 3 13 14
12 4 3 11 9
13 4 4 16 12
14 3 3 9 10
15 3 3 14 13
16 3 4 17 17
17 3 4 10 15

Note: in the results of cluster analysis and the matter-element analysis, 1 refers to huge damage, 2 refers to major damage, 3 refers to moderate damage and 4 refers to small
damage; in the entropy evaluation method and the main component evaluation method, the higher the environmental damage is, the higher the rank will be.

ecological environment; the nos. 5–7 accidents have caused
major damages to the marine ecological environment; the nos. 1,
9, 11, 14–17 accidents have caused moderate damages to the
marine ecological environment and nos. 12 and 13 oil spill
accidents have caused small damages to the marine ecological
environment.
4.2. Comprehensive evaluation of marine ecological environmental
damages caused by oil spill accidents
In this section, we will analyze the examples stated in
Section 4.1 by using some traditional statistical method based on
the optimal cluster number proposed by our method in
Section 4.1. And then, we set up a comprehensive evaluation
model for evaluating marine ecological environmental damages
by respectively using the matter-element analysis method, the
entropy method and the main component method. Evaluation
results of the comprehensive model are shown in Table 4.
From the analysis results shown in Table 4 we can know that
results of mater-element analysis and cluster analysis are very
close. Evaluation results of many accidents are fully consistent and
grades of inconsistent accidents are close as well. Results of
entropy evaluation and main component evaluation are more
detailed and the ranking of damages is given as well.
4.3. Consistency check of comprehensive evaluation of environmental
damage
In order to further prove the rationality of results of the four
evaluation methods, consistency check should be carried out.
Hence, Spearman method is used to analyze relevant evaluation
results. Details are stated in Table 5.
Results of Spearman analysis indicate that results of the four
evaluation methods are strongly correlated with correlation
coefficients all above 0.75 and being notable under the 1%
notability level.
From the above results we can say that the optimal-cluster-based
dynamic evaluation method of environmental damages is rational,
with consistent results to those of traditional evaluation methods.
Additionally, the results obtained from our method reflect the latest
status of environmental damage. Therefore, the method put forward
in this article is practically applicable to dynamically evaluate
environmental damages. Environmental protection departments of
the government are suggested to incorporate the proposed method
into the evaluation of environmental damages. By using the
proposed approach, the government can not only identify the
disaster levels of the to-be-evaluated environmental pollution
events, but also they can get the damage grade of the corresponding
environmental pollution event with the latest data. According to

Table 5
Spearman analysis.

Entropy evaluation method Main component evaluation Matter-element evaluation Optimal cluster
Entropy evaluation method Correlation coefficient 1 .914** .899** .797**
Notability 0 0 0 0
Correlated record number 17 17 17 17
–
Main component evaluation Correlation coefficient .914** 1 .923** .768**
Notability 0 0 0 0
Correlated record number 17 17 17 17
–
Matter-element evaluation Correlation coefficient .899** .923** 1 .858**
Notability 0 0 0 0
Correlated record number 17 17 17 17
–
Optimal cluster Correlation coefficient .797** .768** .858** 1
Notability 0 0 0 0
Correlated record number 17 17 17 17

Note: the notability test is the two-tailed test.

**
p < 0.05
58 T. Yang / Ecological Indicators 51 (2015) 53–58
these evaluation results, the government or the decision making
departments can make more reasonable, suitable and accuracy plans
to take actions to prevent expansion of the pollutions, to rebuild the
damaged ecological environment, and so on.
5. Conclusions
The article put forward a method to dynamically evaluate
corresponding environmental damages based on the optimal
cluster method. Standardization method for treating to-be-
analyzed data of samples is proposed, which can eliminate
effects of dimensions and big differences among data. After
deciding the grade measurement, the system cluster analytic
method is used to analyze the sample data to attain their
classifications and then those data are modified by using the
hierarchical cluster criterion function to get a new criterion
function of the optimal cluster number. Finally, evaluation of
marine ecological environmental damages caused by 17 oil spill
accidents is carried out and the results are compared with those
of traditional evaluation methods. Consistency of those results
is proved. Further, some management suggestions to the
government and the related decision making department are
pointed out.
At least one further research direction can be drawn out from
our research. This article has proposed a new thought to the
researches on dynamic evaluation of environmental damages by
introducing the optimal cluster theory into evaluation. Because
this theory is advantageous in aspects of classification of samples
and decision of optimal cluster numbers, it can be used to
dynamically evaluate other similar disasters by scholars in the
future.
References
Brody, S.D., Peacock, W.G., Gunn, J., 2012. Ecological indicators of flood risk along the
Gulf of Mexico. Ecol. Indic. 18, 493–500.
Burger, J., 2008. Environmental management: integrating ecological evaluation,
remediation, restoration, natural resource damage assessment and long-term
stewardship on contaminated lands. Sci. Total Environ. 400 (1–3), 6–19.
Fallah-Mehdipour, E., Haddad, O.B., Beygi, S., Mariño, M.A., 2011. Effect of utility
function curvature of Young’s bargaining method on the design of WDNs. Water
Resour. Manage. 25 (9), 2197–2218.
Gastineau, P., Taugourdeau, E., 2013. Compensating for environmental damages.
Ecol. Econ. 97, 150–161.
Hanson, D.A., Birtney, E.M., Earie, C.J., et al., 2013. Adapting habitat equivalency
analysis (HEA) to assess environmental loss and compensatory restoration
following severe forest fires. For. Ecol. Manage. 294, 166–177.
Gilbuena Jr., R., Kawamura, Medina, A., Nakagawa, R., Amaguchi, N., 2013.
Environmental impact assessment using a utility-based recursive evidential
reasoning approach for structural flood mitigation measures in Metro Manila
Philippines. J. Environ. Manage. 131, 92–102.
Kennedy, C.J., Cheong, S.M., 2013. Lost ecosystem services as a measure of oil spill
damages: a conceptual analysis of the importance of baselines. J. Environ.
Manage. 128, 43–51.
Li, H., Liu, Y.L., COCA Chen, W.F., 2013. Constructing optimal clustering architecture
to maximize sensor network lifetime. Compu. Commun. 36 (3), 256–268.
Ma, S., Qiang, B., 1988. Environmental management of village and township
enterprises in Guangdong province. Pollution in the Urban Environment Polmet
88 125.
Martin-Ortega, J., Brouwer, R., Aiking, H., 2011. Application of a value-based
equivalency method to assess environmental damage compensation under
the European Environmental Liability Directive. J. Environ. Manage. 92 (6),
1461–1470.
Mulder, W.D., 2013. Optimal clustering in the context of overlapping cluster
analysis. Info. Sci. 223, 56–74.
Rodríguez-Labajos, B., Martínez-Alier, J., 2013. The economics of ecosystems and
biodiversity: recent instances for debate. Conserv. Soc. 11 (4), 326.
Peng, J.F., Song, Y.H., Yuan, P., et al., 2013. A novel identification method of the
environmental risk sources for surface water pollution accidents in chemical
industrial parks. J. Environ. Sci. 25 (7), 1441–1449.
Vaissiere, A.C., Levrel, H., Hilly, C., et al., 2013. Selecting ecological indicators to
compare maintenance costs related to the compensation of damaged ecosystem
services. Ecol. Indic. 29, 255–269.