You are on page 1of 9

Ecological Indicators 7 (2007) 710–718

This article is also available online at:


www.elsevier.com/locate/ecolind

Fuzzy evaluation of water quality classification


Yilmaz Icaga *
Afyon Kocatepe University, Technical Education Faculty, Ahmet Necdet Sezer Campus, 30200 Afyon, Turkey

Received 19 December 2005; received in revised form 1 August 2006; accepted 12 August 2006

Abstract

Traditional classification methods of the water quality parameters use crisp set, and the concentrations values which are close
or far from the limits are considered in same classes. Moreover, usually, several parameters are considered in quality
determination; therefore, differences of the classes of the parameters may be vagueness, especially, in public consideration.
In this study, an index model for quality evaluation of surface water quality classification using fuzzy logic is proposed. In the
method, traditional quality classes are transformed into continuous form and then the concentration values of the different
quality parameters are summed using fuzzy rules, finally, defuzzyfication of this summed values develops the index. An
application of this proposed index model for physical and inorganic chemical parameters in the Eber Lake water (in Turkey) are
studied to demonstrate the practical application and feasibility of this index.
# 2006 Elsevier Ltd. All rights reserved.

Keywords: Water quality; Monitoring; Fuzzy logic; Index; Eber Lake; Modeling

1. Introduction technique may cause a rough and imprecise approach


for data, as in this approach, a parameter being close
Water quality observations have little significance or far from the limit has equal importance for
by themselves. A pollution parameter which has a evaluation of concentration. Secondly, each quality
specific value is usually meaningful only in the parameter may belong to one of four classes. That is,
context of knowledge of natural background levels all of the parameters may not be included in a single
and regulations. Conventional water quality regula- class. These established various quality classes in
tions contain quality classes which use crisp sets, and one sampling location may constitute confusion
the limits between different classes have inherent (ambiguity) for quality definition of that sampling
imprecision (Silvert, 2000). The methods which location.
contain upper and lower limits have two ambiguities. Fuzzy logic can be viewed as a language that allows
Firstly, the traditional water quality evaluation one to translate sophisticated statements from natural
methods use discrete form. This classification language into a mathematical formalism (McNeil and
Thro, 1994). Fuzzy logic can deal with highly
* Tel.: +90 272 228 13 12; fax: +90 272 228 12 35. variable, linguistic, vague and uncertain data or
E-mail address: yicaga@aku.edu.tr. knowledge and, therefore, has the ability to allow

1470-160X/$ – see front matter # 2006 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ecolind.2006.08.002
Y. Icaga / Ecological Indicators 7 (2007) 710–718 711

for a logical, reliable and transparent information support in ecosystem management. They expressed
stream from data collection to data usage in that the application of fuzzy logic seems to be very
environmental applications (Adriaenssens et al., promising in domains such as sustainability, environ-
2004). The concept of fuzzy logic, which is a mental assessment and predictive models (Adriaens-
mathematical discipline based on fuzzy set theory sens et al., 2004). Prato proposed a fuzzy logic
and express multiple levels process among [0, 1] approach for evaluating ecosystem sustainability, and
instead of two levels in classical mathematic (0, 1), stated that the fuzzy logic approach is more
was conceived by Zadeh in 1965. Fuzzy logic appropriate than the conventional crisp sets approach
incorporates a simple, rule-based IF X AND Y THEN to evaluating the strong sustainability of an ecosystem
Z approach to solving a control problem rather than (Prato, 2005). Shen et al. investigated the status of
attempting to model a system mathematically (Sen, combined heavy metal and organo-chlorine pesticide
2001; Elmas, 2003). pollution and evaluated the soil environmental quality
Fuzzy logic could be applied to the development of of the Taihu Lake watershed using a fuzzy compre-
environmental indices in a manner that solves several hensive assessment. The evaluation was carried out in
common problems, including the incompatibility of six steps such as (1) selecting of assessment
observations and the need for implicit value judge- parameters, (2) establishing the membership func-
ments. Fuzzy synthetic evaluation which generally tions, (3) calculating the membership function matrix,
uses a numerical scale to represent water quality and (4) calculating the weights matrix, (5) determination
provides an alternative methodology for aggregating of the fuzzy algorithm and (6) statistical treatment of
the values of the parameters to various quality features data. They stated that the fuzzy comprehensive
have been studied and used in environmental-quality assessment method provides a scientific basis for
evaluation since the 1990s (Ludwig and Tulbure, analyzing and evaluating the environmental quality of
1996; Liou et al., 2003; Liou and Lo, 2005). Recently, soil (Shen et al., 2005). Altunkaynak et al. detected
Lu and Lo developed a multivariable trophic state and removed the linear trend embedded in the
indexing method, for diagnosing water quality and historical records of monthly dissolved oxygen
evaluated this method using fuzzy synthetic evaluation amount and then, they modeled dissolved oxygen
(Lu and Lo, 2002). Haiyan stated that many methods
are tested in trying to assess environmental quality.
These methods can be divided into five general
categories: (1) expert assessment; (2) index assess-
ment; (3) economic analytical method; (4) operational
assessment and (5) fuzzy comprehensive assessment.
In the study, the fuzzy comprehensive assessment is
applied to assess the quality of air, water and soil in
Zhuzhou City, Hunan Province, China based on the
monitoring data of 1997 and National Environmental
Quality Standards of China. The assessment procedure
comprises five steps: (1) select assessment parameters
and establish assessment criteria; (2) establish
membership functions of each assessment parameter
to assessment criteria at each level; (3) substitute the
monitoring data of each assessment parameter at each
monitoring site and national standards into the
membership functions; (4) allocate the weights of
each assessment parameter at each monitoring site to
get a weight matrix; (5) carry out the fuzzy algorithm
(Haiyan, 2002). Adriaenssens et al. reviewed and
assessed applications of fuzzy logic for decision Fig. 1. Flowchart of the index method.
712 Y. Icaga / Ecological Indicators 7 (2007) 710–718

fluctuations in Golden Horn using Takagi–Sugeno 2.1. Fuzzy-rule model in quality assessment
fuzzy logic approach (Altunkaynak et al., 2005).
In this study, the aim is to develop a method based on Conventional classification criteria of water quality
fuzzy logic instead of the conventional crisp classifica- are based either on crisp sets with discontinuous limits
tion method to remove the ambiguities mentioned (e.g. pristine versus polluted), or on continuous
above. In the method, membership functions of the variables whose values are only meaningful to experts
quality parameters and fuzzy rule bases were defined (such as so many ppm of a toxin). Fuzzy sets make it
and then fuzzy logic toolbox of MATLAB1 package possible to combine these approaches (Silvert, 2000).
was used. The method was applied to the 1991–2005 The method developed here contains six steps for
observations of physical and inorganic chemical conventional classification method which contains
parameters of the Eber Lake (Turkey). four classes (Fig. 1):

(Step 1) Determine the quality classes of the para-


2. Methodology meters using observed values.
(Step 2) Arrange the parameters for their classes to
Fuzzy logic contains, usually, fuzzification, applica- obtain four groups.
tion of the rule base to fuzzy data, inference of fuzzy (Step 3) Use membership functions to standardize the
result and defuzzification of fuzzy result stages. natural measurement scales of the quality
Fuzzification is a process that transforms the observed parameter into a measurement of the quality
(real) data to fuzzy form using membership function degree (membership grade). In this step, four
that defined for problem features. The rule base defines membership functions are used. Namely;

8
>
<1 for x < a
mf 1 ¼ 1  ½ðx  aÞ=ðb1  aÞ for a  x < b1 (1)
>
:
0 for otherwise
8
> ðx  aÞ=ðb1  aÞ for a  x < b1
>
>
< 1 for b1  x < b2
mf 2 ¼ (2)
>
> 1  ½ðx  b2 Þ=ðc1  b2 Þ for b2  x < c1
>
:
0 for otherwise
8
> ðx  b2 Þ=ðb  c1 Þ for b  x < c1
>
>
<1 for c1  x < c2
mf 3 ¼ (3)
>
> 1  ½ðx  c2 Þ=ðd  c2 Þ for c2  x < d
>
:
0 for otherwise
8
< 1  ½ðx  c2 Þ=ðd  c2 Þ for c2 < x  d
>
mf 4 ¼ 1 for d < x (4)
>
:
0 for otherwise

the relationship among the membership functions and where mfi is the ist membership function, x
the form of the result membership function. Defuzzi- the observed value; a, b, c, d are the limits of
fication provides the real value from the result the related membership functions (Tables 1
membership function (McNeil and Thro, 1994; Sen, and 2); in case the membership function is in
2001; Elmas, 2003; Bandemer and Gottwald, 1996). trapezoid shape then the b1 and c1 are the
Y. Icaga / Ecological Indicators 7 (2007) 710–718 713

first top levels of the trapezoids and b2 and c2 physical and inorganic chemical parameters, organic
are the second top levels of the trapezoids parameters, inorganic pollution parameters and
otherwise b1 = b2 and c1 = c2. bacteriologic parameters. Each of these groups has
(Step 4) Use four rule bases: four limits for each parameter. The limits of the

if QP1 ¼I or QP2 ¼I or ... or QPN ¼I then QP ¼ I


if QP1 ¼ II or QP2 ¼ II or ... or QPN ¼ II then QP ¼ II
(5)
if QP1 ¼ III or QP2 ¼ III or ... or QPN ¼ III then QP ¼ III
if QP1 ¼ IV or QP2 ¼ IV or ... or QPN ¼ IV then QP ¼ IV

where QPi is the ist quality parameter; I, II, physical and inorganic chemical parameters are
III, IV are the quality classes in traditional presented in Table 1.
classification; N is the number of quality Observations used in this study comprise eleven of
parameters. In the rule bases the ‘‘or’’ 13 quality parameters given in the Water Pollution
operators are used to obtain maximum Control Regulations (Table 1) (Official paper, 1998;
values. Icaga et al., 2006). In 1991, the State Institution for
(Step 5) Use the fuzzy algorithm. In fuzzy algorithm, Hydraulic Works initialized observations of water
the Mamdani approach is used. Fuzzy quality data in the six sampling locations of the Eber
inferences of the groups are determined Lake, periodically. The observed parameters (the
using grades of membership functions of the parameters were written without their valence values
parameters. for easy edition) in the basin are temperature (T) (8C),
(Step 6) Defuzzificate the four fuzzy inferences of the pH, dissolved oxygen (DO) (g/m3), chloride (Cl) (g/
groups to obtain an index number between 0 m3), sulphate (SO4) (g/m3), ammonia (NH3) (g/m3),
and100 intervals using Centroid methods nitrite (NO2) (g/m3), nitrate (NO3) (g/m3), total
which represents center of gravity of the dissolved solid (TDS) (g/m3), color (Pt-co) and
membership function (for details about fuzzy sodium (Na) (g/m3).
logic see (Bandemer and Gottwald, 1996).
3.2. Results

3. Case study In Table 2 the mean values of the data are


presented. The quality classes of the parameters are
3.1. Study area given in Fig. 2. The fuzzy logic calculation steps
presented in Fig. 1 were applied to the data in the
The Eber Lake is a fresh water lake, located in the following order:
inner west of the Turkey. The area of the lake is
156 km2, its volume is about 333 hm3, and the (Step 1) Quality classes of the parameters were
maximum depth is 4.6 m. The water level changes determined using conventional classification
with season and years, and the flood water is (Fig. 2).
discharged to the Aksehir Lake. The most important (Step 2) The parameters were grouped for their
source of the Eber Lake is Akarcay stream. The lake classes.
has rich nutrient content ecologically. The diversities (Step 3) Used membership functions, which their
of animals and plants are extremely rich. More than shapes were in triangle and trapezoid form,
200 bird species, with more than 30,000 individuals, were derived form the limits given in
are established in the Eber Lake (TMEF). Table 1. The limits in Table 1 were
In Turkey, the water quality parameters have been determined as interval (transition) location.
defined in the scope of water quality criteria as Middle values of the limits determined as the
714 Y. Icaga / Ecological Indicators 7 (2007) 710–718

Table 1
Quality classes for physical and inorganic chemical parameters of inside water resources and observed values (Official paper, 1998; Icaga et al.,
2006)
Limits of quality classes Observations
I II III IV
Temperature (T) (8C) 25 25 30 >30 16.75
pH 6.5–8.5 6.5–8.5 6–9 <6–9> 8.40
Dissolved oxygen (DO) (g/m3) 8 6 3 <3 8.53
Oxygen saturate (OS) (g/m3) 90 70 40 <40 No data
Chloride (Cl) (g/m3) 25 200 400 >400 145.74
Sulphat (SO4) (g/m3) 200 200 400 >400 499.33
Ammonia (NH3) (g/m3) 0.2 1 2 >2 2.108
Nitrite (NO2) (g/m3) 0.002 0.01 0.05 >0.05 0.026
Nitrate (NO3) (g/m3) 5 10 20 >20 0.930
Total phosphors (g/m3) 0.02 0.16 0.65 >0.65 No data
Total dissolved solid (TDS) (g/m3) 500 1500 5000 >5000 1329.08
Color (Pt-co unit) 5 50 300 >300 21.82
Sodium (Na) (g/m3) 125 125 250 >250 235.88

Table 2
The limits of the membership functions
The parameters of the membership functions
a b c d
a b a b
b1 b2 c1 c2
Temperature (T) (8C) 17.5 22.5 27.5 32.5
pH  7.5 7. 5 7.75 8.75 9.25
pH < 7.5 5.75 6.25 6.75 7.5
Dissolved oxygen (DO) (g/m3)c 9 7 4.5 1.5
Chloride (Cl) (g/m3) 0 50 100 300 500
Sulphat (SO4) (g/m3) 50 150 250 350 450
Ammonia (NH3) (g/m3) 0 0.4 1.5 2.5
Nitrite (NO2) (g/m3) 0 0.004 0.03 0.07
Nitrate (NO3) (g/m3) 2.5 7.5 15 25
Total dissolved Solid (TDS) (g/m3) 0 1000 3250 6250
Color (Pt-co unit) 0 27.5 175 425
Sodium (Na) (g/m3) 31.25 93.75 156.3 218.8 281.3
Output membership function 12.5 37.5 62.5 87.5
a
First upper corner of the trapezoidal membership function.
b
Second upper corner of the trapezoidal membership function.
c
The number membership functions in reverse order (see Fig. 6).

maximum grade of the membership func- val with class III. The parameters of the
tions. Some parameters in the control membership functions were given in Table 2.
regulations have one limit for two classes. For every class of water quality parameters
This limit has been divided into two levels. and also for the output, a triangular or a
For example, temperature parameter has one trapezoidal membership function was
limit for the class I and class II. So, two new defined and the plots of the functions are
limits were generated considering the inter- given in Figs. 3–15.
Y. Icaga / Ecological Indicators 7 (2007) 710–718 715

Fig. 5. Membership functions of pH (for pH < 7.5).

Fig. 2. Quality classes of physical and inorganic chemical para- 4. Discussion


meters in the Eber Lake. Aim of the study was to remove two ambiguities
observed in the data. Firstly, concentration levels of
the parameters near to class limits were considered
sensitively. Just as an example, the observed value of
pH is 8.40, which is near the limit, and this value is in
class II. However, a 1.1% increment (that is 0.10) of
the parameter concentration changes the quality
classes from class II to class III.
For the second ambiguity following statements can
be deducted: Fig. 2 shows us that there is not
Fig. 3. Membership functions of temperature. homogeneity among the parameters for the quality
classes and, there is at least one parameter in each
(Step 4) In this stage, the rule bases were arranged quality class of conventional method. That is, the
using the quality classes of the parameters: parameters T, DO, NO3 are in class I; pH, Cl, TDS and

IF T ¼ I or DO ¼ I or NO3 ¼ I THEN Output ¼ I


IF pH ¼ II or Cl ¼ II or TDS ¼ II or color ¼ II THEN Output ¼ II
IF NO2 ¼ III or Na ¼ III THEN Output ¼ III
IF SO4 ¼ IV or NH3 ¼ IV THEN Output ¼ IV

(Step 5) The fuzzy inferences of the four groups were color are in class II; NO2 and Na are in class III; SO4
calculated. and NH3 is in class IV. The parameters constitute
(Step 6) Lastly, the inferences were defuzzificated, ambiguity on assessment of water quality of the lake
and so, the index of the lake was found as due to the various quality classes. Therefore, two
41.2. ambiguities which were defined above in water quality

Fig. 4. Membership functions of pH (for pH  7.5). Fig. 6. Membership functions of DO.


716 Y. Icaga / Ecological Indicators 7 (2007) 710–718

Fig. 7. Membership functions of Cl. Fig. 11. Membership functions of NO3.

Fig. 8. Membership functions of SO4. Fig. 12. Membership functions of TDS.

Fuzzy logic results produce 41.2 which shows that


the quality degree in the sampling location and can be
considered as an index. If one wants to express this
index value in the present classification, this value
reflects the class II. Therefore, both in index and in
present classical classification, the water quality was
stated in an easy understandable way.
The proposed method in this study was applied
only physical and inorganic chemical parameters. The
Fig. 9. Membership functions of NH3. method decreased the 11 quality values to one
synthetic value and also removed the impreciseness
in the limits of the quality classes. This method can
assessment were resulted in these data. Namely, the also be used for the other parameters like organic
limits of the classes have imprecisions and should be parameters, inorganic pollution parameters and
considered in fuzzy logic, and also the water quality bacteriologic parameters. Besides this, it can be
assessment in the parameters has different quality easily understood that how complex results can be
classes and they have no homogeny. confronted in a classification analysis with conven-

Fig. 10. Membership functions of NO2. Fig. 13. Membership functions of Color.
Y. Icaga / Ecological Indicators 7 (2007) 710–718 717

form and were combined to one index value was


obtained.
More accurate information may be obtained by
using continuous form which is obtained by fuzzy
logic. Conceptually, the use of fuzzy logic is simple,
but the power of the method used here comes from the
ability to integrate different kinds of quality observa-
tions. However, incommensurable effects were not
Fig. 14. Membership functions of Na. used but it is thought that the incommensurable effects
can be considered in this method. This method may
therefore be appropriate for the assessment water
pollution.

References

Adriaenssens, V., De Baets, B., Goethals, P.L.M., De Pauw, N.,


2004. Fuzzy rule-based models for decision support in ecosys-
tem Management. Sci. Total Environ. 319, 1–12.
Fig. 15. Membership functions of output. Altunkaynak, A., Ozger, M., Cakmakci, M., 2005. Fuzzy logic
modeling of the dissolved oxygen fluctuations in Golden Horn.
Ecol. Model. 189, 436–446.
tional method for all parameters. However, an index is Bandemer, H., Gottwald, S., 1996. Fuzzy Sets, Fuzzy Logic Fuzzy
a more precise quality indicator then the conventional Methods with Appications. John Wiley & Sons, Chichester, p.
239.
method as it associates different types of observa- Elmas, C., 2003. Fuzzy logic inspections (theory, application,
tions. Neural Fuzzy Logic), Seckin, Ankara, p. 230 (in Turkish).
Haiyan, W., 2002. Assessment and prediction of overall environ-
mental quality of Zhuzhou City, Hunan Province, China. J.
5. Conclusions Environ. Manage. 66, 329–340.
Icaga, Y., Bostanoglu, Y., Kahraman, E., 2006. Akarçay su kalitesi
istatistikleri. Technol. Res.: EJCT 2 (1), 43–50.
In the study, firstly, quality limits of the traditional Liou, S.M., Lo, S.L., Hu, C.Y., 2003. Application of two-stage fuzzy
classification were considered in continuous form set theory to river quality evaluation in Taiwan Source. Water
using fuzzy logic. Therefore, the evaluation sensitivity Res. 37 (6), 1406–1416.
Liou, Y.T., Lo, S.L., 2005. A fuzzy index model for trophic
of concentration is increased. Secondly, various
status evaluation of reservoir waters. Water Res. 39, 1415–
parameters are, usually, in different quality classes 1423.
and this may confuse non experts on the subject. Lu, R.S., Lo, S.L., 2002. Diagnosing reservoir water quality using
Therefore, these classes were combined with fuzzy self-organizing maps and fuzzy theory. Water Res. 36, 2265–
logic and one index value was obtained to represent 2274.
quality classes of all parameters in order to make water Ludwig, B., Tulbure, I., 1996. Contributions to an aggregated
environmental pollution index. In: Proceedings of the Inter-
quality assessment more comprehensible especially in society Energy Conversion Engineering Conference, vol. 3. pp.
public consideration. 2144–2149.
The physical and inorganic chemical parameter MATLAB1 7.0; Packaged software for technical computing,
observations in one sampling location of the Eber Release 14, The Math Works, Inc.
McNeil, F.M., Thro, E., 1994. Fuzzy Logic A Practical Approach.
Lake were used. The parameters were classified in
AP Professional, Boston, p. 294.
traditional method. There were four quality classes in Official paper, 1998. Water pollution control regulations, no. 19919,
the parameters. These observations were also con- 13–74 (in Turkish)
sidered in the method proposed in this study. Firstly, Prato, T., 2005. A fuzzy logic approach for evaluating ecosystem
the traditional classes were transferred to continuous sustainability. Ecol. Model. 187, 361–368.
718 Y. Icaga / Ecological Indicators 7 (2007) 710–718

Sen, Z., 2001. Fuzzy logic and modeling principles. Bilge Kultur Silvert, W., 2000. Fuzzy indices of environmental conditions. Ecol.
Sanat, İstanbul, p. 174 (in Turkish). Model. 130, 111–119.
Shen, G., Lu, Y., Wang, M., Sun, Y., 2005. Status and fuzzy TMEF (Turkey Ministry of Environment and Forestry), 2006. Eber
comprehensive assessment of combined heavy metal and Gölü (in Turkish), URL: http://www.cevreorman.gov.tr/sulak/
organo-chlorine pesticide pollution in the Taihu Lake region sulakalan/eber.htm.
of China. J. Environ. Manage. 76, 355–362.

You might also like