Fuzzy expert system for the detection of episodes of

poor water quality through continuous measurement

Cecilio Anguloa,∗, Joan Cabestanya, Pablo Rodŕıguezb, Montserrat Batlleb,
Antonio Gonzálezb, Sergio de Camposb

aUniversitat Politècnica de Catalunya, UPC-Barcelona Tech, Vilanova i la Geltrú, Spain
bAdasa Sistemas S.A.U., Barcelona, Spain

Abstract

In order to prevent and reduce water pollution, promote a sustainable use,
protect the environment and enhance the status of aquatic ecosystems, this
article deals with the application of advanced mathematical techniques de-
signed to aid in the management of records of different water quality moni-
toring networks. These studies include the development of a software tool for
decision support, based on the application of fuzzy logic techniques, which
can indicate water quality episodes from the behaviour of variables mea-
sured at continuous automatic water control networks. Using a few physical-
chemical variables recorded continuously, the expert system is able to obtain
water quality phenomena indicators, which can be associated, with a high
probability of cause-effect relationship, with human pressure on the water
environment, such as urban discharges or diffuse agricultural pollution. In
this sense, at the proposed expert system, automatic water quality control
networks complement manual sampling of official administrative networks
and laboratory analysis, providing information related to specific events (dis-
charges) or continuous processes (eutrophication, fish risk) which can hardly
be detected by discrete sampling.

Keywords: Water quality system, Automated measurement networks,
Fuzzy logic, Fuzzy inference system, Guadiana river

∗Neàpolis Building, Rambla de l’Exposició 62, 08800 Vilanova i la Geltrú, Spain. Tel:
+34 938967273; Fax: +34 938967200

Email address: cecilio.angulo@upc.edu (Cecilio Angulo)

Preprint submitted to Expert Systems with Applications June 7, 2010



1. Introduction

On October 23th, 2000, the “Directive 2000/60/EC of the European Par-
liament and of the Council establishing a framework for the Community
action in the field of water policy” or, in short, the EU Water Framework
Directive (European Parliament, 2000) was finally adopted. The Directive
is about the organization of water management, in order to prevent and re-
duce pollution, promote sustainable water use, protect the environment and
enhance the status of aquatic ecosystems.

As a starting point in the practice of the policy in Spain, tasks for the
characterization of anthropogenic impact on the waters were carried out,
as well as the diagnosis of their condition based on monitored indicators.
Application of the Directive led to the execution of an analysis of the char-
acteristics of each river basin, and a study of the impact of human activity
on the water.

Based on the results of such initial analysis as well as determining fac-
tors such as cost-benefit analysis, citizen participation and others, each river
basin’s management unit should be able to develop a management plan and
a program of measures. The objectives of this plan would prevent deteriora-
tion, enhance and restore the status of surface water bodies, ensuring that
they are in good ecological and chemical status, and reduce pollution due to
discharges and emissions of hazardous substances.

Moreover, the deployment of the stations of the Automated Information
Water Quality System (SAICA, in Spanish) has allowed to obtain automated
and continuous information from the state of rivers (Serramià, 2005). Auto-
mated information is different from those usually collected through manual
measurements. Unlike the latter, which are timely and accurate measures,
the SAICA ones present the advantages of immediacy and continuity. How-
ever, they involve the management of large quantities of records spread across
multiple variables that are supposed to be less precise.

This article presents work developed through the Spanish project ECOW-
ATCH in the context of implementing the EU Water Framework Directive.
It deals with the application of advanced mathematical techniques designed
to aid in the management of records from different observational networks of
water quality. Studies and developments performed in the project ECOW-
ATCH include the development of a software tool for decision support, based
on the application of fuzzy logic techniques, which can detect quality episodes
from the behavior of continuous variables measured in SAICA networks.

2



The developed expert system is able to generate, in real time, a set of
combined indicators of water quality based on:

• data from SAICA automated networks

• expert knowledge about episodes of poor water quality expressed as
rules

whose purpose is to detect two different types of events:

• ad hoc and spurious events, like urban discharges or caused by a waste
water treatment plant (WWTP),

• long and cumulative over time events, like episodes of eutrophication
and, as a novelty, fish risk, i.e. risky environmental conditions for fish
communities

The working assumption along this project is that expert definition of
continuous indicators will provide information on the evolution of the water
quality of a river basin allowing early detection of abnormal events. These
indicators will anticipate and complete the results obtained with hand-made
programmed sampling. In addition, this form of dual management:

• will increase knowledge about the river basin, by summarizing in syn-
thetic values the information contained in a set of complex variables

• justifies the operation and maintenance costs of continuous automatic
monitoring networks, currently used for warning and inspection pur-
poses

Automatic monitoring networks will supply an added value as knowledge
generators about the environment and can help assess the effectiveness and
appropriateness of the actions taking place on the basin or specific parts of the
same: introducing a management system, construction of a water treatment
infrastructure or others.

Based on the working hypothesis and with the aim of combining expert
knowledge with information collected by SAICA network, the methodology
depicted in Figure 1 has been used as project development script. As shown
therein, the generation of indicators by the expert system is based on the
expression of empirical expert rules in the form of fuzzy logic. Such expert
fuzzy rules are constructed from a set of examples that allow the formulation

3



of the phenomena by the experts. The acquisition of this knowledge through
the use of fuzzy logic will allow to design an expert system able to generalize
to new situations.

Figure 1: Development process for setting-up fuzzy logic rules.

2. Selecting poor water quality episodes

The Spanish side of the Guadiana river basin has been chosen as the study
area for this work, due to the availability of information records collected by
several automatic measurement stations belonging to the Guadiana Hydro-
graphic Confederation’s SAICA network. Figure 2(a) shows the location of
the basin and the selected stations are depicted in Figure 2(b).

2.1. Episodes and related variables

The expert system to be developed will focus on the detection of three
types of episodes: (i) discharges to the environment, (ii) eutrophication and,
experimentally, (iii) fish risk. The determination of quality variables to be
used as inputs for each kind of episode to assess and their combined impact
has been completed with the collaboration of experts, who have characterized
a series of remarkable events, which have been used to calibrate the developed
methods.

4



(a) Guadiana River basin. (b) Stations in the Guadiana River.

Figure 2: Location of selected stations in the Guadiana River.

The five variables extracted from expert knowledge to characterize episodes
have been: pH, conductivity, turbidity, ammonium concentration and dis-
solved oxygen (DO) concentration.

Relevant cases corresponding to the studied episodes should be selected
for training and development of automated fuzzy inference systems. In the
following, episodes characterizing each type of analyzed event are discussed,
along with a detailed description in terms of the consulted experts.

2.2. Episodes of Urban or WWTP discharge

The two episodes of water discharge shown below correspond to data from
the Guadajira automatic station, located downstream of a WWTP.

2.2.1. Episode 1. Guadajira, January 2009

The values of the variables selected by the experts for the duration of the
episode are shown in Figure 3. Two different events can be distinguished,
one of them is starting on January 22th, and another at the beginning of the
27th, marked with two separate rectangles.

An explanation for the first event is a possible malfunction in the process
of nitrification - denitrification of the upstream wastewater treatment plant,
which would justify the increase of ammonium, as well as in the phosphate
removal, while there is a decrease in the dissolved oxygen concentration. For
the second event, aeration might not work properly in the WWTP, so no
removal of phosphorus and ammonium exists, while there is an increase in

5



Figure 3: Measured variables for Episode 1 (discharge).

organic matter, which is reflected in the decrease of dissolved oxygen. Both
cases are similar, the second one having a largest pollution load than the
first, although it could also be a direct discharge.

This kind of situation is not unusual in this particular WWTP. A planned
action exists for this specific WWTP in order to enlarge it to provide sufficient
treatment capabilities for current needs. The proposed expert system would
serve, in this case, as a tool for assessing the effectiveness and appropriateness
of the actions currently being undertaken at the WWTP.

2.2.2. Episode 2. Guadajira, September 2008

In Figure 4 the values of the variables during the analyzed event are
depicted. It can be seen that the event starts in the evening of September
22th, as marked with a rectangle.

The increased presence of ammonium indicates that the process of nitri-
fication - denitrification has not been completed, therefore the cause is in
the WWTP. Moreover, there has neither been degradation of organic mat-
ter before the discharge, since values for the dissolved oxygen concentration
become null and turbidity increases due to the presence of organic matter.

6



Figure 4: Measured variables for Episode 2 (discharge).

This episode is related to a direct urban discharge, from either the WWTP
itself, or from some of the olive-related industries located in the area that
are not networked to the sewerage system which serves the WWTP, and that
are in the phase of requirements for implementation of appropriate treatment
systems.

2.3. Episode of Eutrophication

A process of eutrophication is an increase in the trophic status of water
caused by nutrient enrichment (Smith et al., 1999). In the limit, the cumula-
tive process of eutrophication can lead to a biological impact as it comes to
situations of partial (night) or permanent anoxia (no oxygen), which could
end in itself with fish life. The selected episode to be used for the expert
system is shown in the next section.

2.3.1. Episode 3. Benavides, March 2007

A process of natural photosynthesis is going on: daytime primary pro-
ducers consume CO2 and produce oxygen, thus decreasing H

+ concentration,
and therefore pH increases. At night CO2 concentration increases, increasing
H+, and low pH. As the process continues, the possible presence of N and
P nutrients (not measured) causes biomass growth and increases production

7



Figure 5: Measured variables for Episode 3 (eutrophication and fish risk).

/ consumption of oxygen, which is manifested in the amplitude of oscillation
in the daily cycle of this element. While N and P values are in excess the
process will increase.

From an initial state of eutrophication, low flow does not favor dilution,
and the increase of conductivity may indicate a concentration of salts and
increased nutrient concentrations (nitrates, phosphates...) that are a limit-
ing factor for algae and macrophytes, favoring the uncontrolled growth of
biomass. The system is no longer sustainable as the amount of biomass
to degrade exceeds the system’s capacity, hence oxygen concentration de-
creases until eventual anoxia situations. Values for quality variables during
the episode are shown in Figure 5.

2.4. Episode of Fish risk

According to the experience of operation managers, during the advanced
stage of eutrophication an increase in average pH is observed. In these cir-
cumstances, if a discharge of ammonium in a basic medium exists, it is trans-
formed into ammonia, a highly toxic product and lethal to fish fauna, which
was already under pressure due to periods of partial anoxia.

Information about fish risk is useful for operation managers as an aid to

8



decision making for preventive action. In entering a risk area, when there is
a capacity to act, dilution can be forced by releasing water from reservoirs to
help breaking the cycle of eutrophication. An episode from a water station
on the River Guadiana has been selected and is shown in the next section.

2.4.1. Episode 3. Benavides, March 2007

Ammonium (NH+4 ) is in the water in steady equilibrium with ammonia
(NH3), highly toxic (NH3 +H

+ ↔ NH+4 ). Faced with a discharge of ammo-
nium in the river, a small increase in pH (H+ decrease) moves the equilibrium
towards the formation of NH3. While ammonium can be easily assimilated,
ammonia is lethal to living beings. The situation can be particularly danger-
ous on this zone of the river, which consists on a wide and clear channel with
shallow waters, when eutrophication occurs. Both contributing factors lead
to a strong proliferation of phytoplankton, which during the daily cycle cause
significant variations in pH. During an episode with high levels of ammonium
that can not be neutralized by oxidation, NH3 lethal concentrations to the
fish stock can be reached.

In this episode, the average pH values grow up close to 9, so a small
discharge of ammonium is dangerous to the fish fauna, and a large one,
lethal. Figure 5 shows the value for quality variables during the episode.

3. Fuzzy inference systems developed for the expert system

The set of examples of abnormal events to be used for the definition and
development of the expert system have been established. The next stage
in the methodology is the automation, by the working group, of the steps
to generate the expert system (see Figure 1). A first step, which consumes
a significant amount of time if it is not well automated, is the inspection
and validation of the database provided by the SAICA network’s stations.
In particular, measurement errors for out of range and non-present values
in some variables should be considered. Standardized solutions have been
adopted: absent values have been filled by linear regression and outliers have
been erased and replaced with mean data from previous and consequent
samples. Experts have been paying particular attention to some little samples
difficult to fix automatically.

3.1. Description

Fuzzy inference systems are a widespread method in the treatment of
information for the generation of expert systems in monitoring the water

9



quality (Nasiri et al., 2007; Lermontov et al., 2009; Hatzikos et al., 2009).
Roughly speaking, a fuzzy inference system is an algorithm able to convert
a strategy or set of linguistic rules in an automated strategy. Applications
can be designed with fuzzy logic so that systems and processes respond with
greater knowledge to imprecision, which seeks to mimic human behavior.

Our approach is focused on the automated generation of fuzzy rules from
recorded data according to a pre-defined structure of rules or strategy de-
termined by human experts. Once this strategy is provided, some questions
must be answered for the generation of the automated inference system. In
particular:

1. How are membership functions for both input and output variables
set-up?

2. Which labels for the output variable are assigned for each rule?

3. How are linguistic labels determined for the defuzzified output?

Before developing the automated system, effectiveness of using fuzzy in-
ference systems was firstly validated by manually creating an expert system
prototype, which was tested on the two episodes of urban discharge in Guada-
jira’s SAICA station, in January 2009 and September 2008. As a result, for
the first of the episodes, a set of 54 basic rules was designed. One example
of them is following:

54. IF (IncConduct is Negative) and (IncOD is Zero) and

(IncAmmonium is Negative) and (IncPhosphates is Negative) THEN

(UrbanDischarge is Normal)

For the second example, turbidity measures are further provided, so the
initial knowledge base is completed with the information of whether or not
there is increased turbidity. This resulted in 162 rules. One example is
following:

1. IF (IncConduct is Positive) and (IncOD is Negative)

and (IncTurbidity is Positive) and (IncAmmonium is Positive)

and (mOD is Low) THEN (UrbanDischarge is Evidences)

After positive testing of manually generated fuzzy inference systems for
these two examples, it was time to generalize, i.e. to develop an automated
procedure to automatically obtain fuzzy rules according to provided data.

10



3.2. General Application

The continuous data obtained through automated networks in the Guadi-
ana river Hydrographic Confederation (CHG) have allowed characterization
of different episodes in its basin, which experts have sought to associate the
possible cause of. As an instance, in Figure 3 the expert has been able to
find out two episodes of increased ammonium matching a decrease in oxy-
gen, a slight decrease in pH and a slight increase in conductivity. In turn, a
discharge of organic origin has been assigned as a possible cause.

This type of information is quite important to perform a systematic study
of water quality, so it is very important to have the support of technical
experts who can discern this kind of behavior and match it with real expe-
riences. The creation of an expert system capable of collecting, objectifying
and quantifying this experience, can be made more transparent through an
inference system based on fuzzy rules. The fuzzy inference systems should
allow the identification of combined indices of automatic settings that are
representative of specific events or indicators of possible trends in the evolu-
tion of the pressures. Figure 6 shows the two types of application processes
that have been followed for the development and calibration of the inference
system, distinguishing between an automatic formulation of rules once the
principles of phenomena are known, and a manual formulation for the risk
fish phenomenon due to its special nature.

3.2.1. Expert strategy formulation

As a result of the analysis of the information gathered by experts, a
strategy formulation has been provided for the three types of events to be
detected.

The resulting algorithm for the detection of discharge episodes is the
concatenation of three fuzzy inference systems, resulting in a value between
0 and 1 drawn from the linguistic labels “Normal”, “Evidence of discharge”,
“Slight discharge” and “Serious discharge” about the output variable (Figure
7). Features were selected among incremental and mean values of the original
variables, according to expert advice. A point to be studied later is how to
best combine the outputs of the three inference systems, either by nesting
them or by weighting the outputs.

For the detection of eutrophication phenomena, the resulting algorithm
is composed of a single fuzzy inference system, resulting in a value between
0 and 1 extracted from labels “Normal”, “Natural photosynthesis”, “Slight
eutrophication” and “Advanced eutrophication” in the output. A diagram of

11



Figure 6: Algorithm calibration process.

Figure 7: Fuzzy inference system strategy for the ‘discharge indicator’.

12



Figure 8: Strategy for the ‘eutrophication indicator’.

the obtained algorithm is shown in Figure 8. In this case, since eutrophication
is a process lasting for several days and depending on the day/night cycle,
features were designed using a long temporal window.

Finally, the algorithm for detecting fish risk situations consists of a single
fuzzy inference system, resulting in a value between 0 and 1 drawn from the
values of membership functions on linguistic labels “Normal”, “Possible risk”,
“Moderate risk”, “High risk” and “Extreme risk” in the output. In this case,
membership functions for the input variables have been manually designed.
Representations of the membership functions, as designed by experts, have
been depicted in Figure 9.

3.2.2. Automated design

Once strategies have been provided by the experts, membership functions
for the input and output variables and labels for the output in the rules must
be calculated in an automated way from the SAICA network data. Later,
they should be tuned or scaled so that the response is adjusted to the expert
opinion on the known cases.

Input variables’ membership functions. For the input variables, membership
functions have been designed according to the following constraints:

• three membership functions for each input variable

13



Figure 9: Fuzzy algorithm strategy for the ‘fish risk indicator’.

• the universe of discourse is defined by the database ranges and symme-
try

• membership functions are gaussian functions centered on both bounds
of the range and its middle

• gaussian functions’ width is a compromise between data quartiles and
normalization to the unity

As an illustrative example, membership functions for the ‘DO increment’
variable are shown in Figure 10.

For the ‘fish risk indicator’ case, the expert-designed membership func-
tions for the input variables (depicted in Figure 9) have been translated to
those shown in Figure 11.

Output variable’s membership functions. For the output variable, member-
ship functions have been designed according to the following constraints:

• usual equally distributed triangular membership functions have been
employed with the universe of discourse in the range [0, 1], for the final
output variable

14



Figure 10: Membership functions for the variable ‘DO increment’.

• for the concatenated fuzzy inference systems in the ‘discharge indi-
cator’, the same distribution has been chosen with four membership
functions

Output variable’s labels for each rule. Labels for the output variable for each
rule are also automatically associated by considering the strength of the input
variables in the rule’s precedent, and its either direct or reverse relationship
with the rule’s consequent. An easy linear formulation has been employed to
assign a label to the output variable.

Linguistic labels for the defuzzified output. Tresholds used to assign linguis-
tic labels to the indicators have been calculated after the results on known
episodes have been analyzed. Output normalization to the range [0, 1] could
have been performed. However we are interested on generalization to other
basins and rivers, hence it was determined not to scale it. Output normaliza-
tion has been delayed for future studies on data coming from further Spanish
river basins.

4. Experimentation and result analysis

A MATLAB/GUI (Graphical User Interface) based tool has been devel-
oped as a fuzzy expert system for the detection of episodes of poor water

15



(a) Fuzzy inference system for the ‘fish risk indicator’.

(b) Mean pH. (c) Mean DO. (d) Mean Ammonium.

Figure 11: Strategy and input membership functions for the ‘Fish risk indicator’.

quality (discharge, eutrophication and fish risk) through continuous mea-
surement. The prototype has been evaluated on the SAICA network data
provided by the river Guadiana Hydrographic Confederation and obtained
results have been analyzed. An example of the GUI-based user-friendly in-
terface is illustrated in Figure 12.

The indicator of urban or WWTP discharge (‘discharge indicator’) has
been specifically calibrated for the Guadajira station and for episodes 1 and
2. It should be adjusted if implemented in other locations. This process
could be performed with a set of selected episodes or with a long series of
data.

The ‘eutrophication indicator’ has been tuned for episode 3, and it has
been directly applied to other stations (in the same basin or a different one)
without ulterior adjustement.

Finally, the ‘fish risk indicator’ is an absolute indicator, and no tuning
process was performed in any case. Consequently, it can be directly applied
to other stations or basins.

16



Figure 12: MATLAB/GUI prototype’s example.

Below are the results of the implementation of the algorithms for calcu-
lating event indicators.

4.1. Urban or WWTP discharge

Figure 13 (top) shows the results for the ‘discharge indicator’ for Episode
1. It can be noted that the first event has been characterized as “Evidence
of discharge”, while the second has been labeled as “Serious discharge”. The
indicator has identified yet another possible event at the beginning of the 24th
day, mainly due to an increase in ammonium concentration and a lowering
of dissolved oxygen, but not important enough to become an “Evidence of
discharge”.

Figure 13 (bottom) shows the results for the ‘discharge indicator’ for
Episode 2. In this case the indicator has a base value slightly below the
threshold of an “Evidence of discharge”, so sometimes it is occasionally over-
passed.

The event on the afternoon of the 22nd day is properly detected and
characterized as “Serious discharge”. And just at the beginning of the 24th
day there is another event that is detected at the edge of being considered
“Slight discharge”. An inspection of the original data does not show that it
is a new event, but just a spike in long event conditions detected in the first
place.

The results suggest the use of the ‘discharge indicator’ thresholds shown
in Table 1 to characterize detected events more easily.

17



Figure 13: ‘Discharge indicator’ results for Episodes 1 and 2.

Discharge “Normal” “Evidence” “Slight” “Serious”
Range < 0.4 0.4 − 0.6 0.6 − 0.8 > 0.8

Table 1: Thresholds for the ‘discharge indicator’.

18



Figure 14: ‘Eutrophication indicator’ and ‘Fish risk indicator’ results for Episode 3.

Eutrophia “Normal” “Photosynthesis” “Slight” “Advanced”
Range < 0.2 0.2 − 0.3 0.3 − 0.4 > 0.4

Table 2: Thresholds for the ‘eutrophication indicator’.

In measuring stations located downstream of discharges, at most “Ev-
idence of discharge” values were only detected. This makes sense because
when water flowed into the main channel of the Guadiana river, dilution
occurred and decreased the effects of the discharge.

4.2. Eutrophication

The ‘eutrophication indicator’ for Episode 3 enters a state of “Natural
photosynthesis” from the 9th day and gradually grows up into “Slight eu-
trophication” in the middle of the study period. At this moment, the state
moves rapidly to “Advanced eutrophication”, where it remains until the 30th
day. Since then decreases progressively, as shown in Figure 14 (top).

The results in the ‘eutrophication indicator’ suggest using the thresholds
defined in Table 2 to characterize the detected events.

It can be noted that the obtained values for this indicator have been lower

19



Fish risk “Normal” “Possible” “Moderate” “High” “Extreme”
Range < 0.2 0.2 − 0.3 0.3 − 0.4 0.4 − 05 > 0.5

Table 3: Thresholds for the ‘fish risk indicator’.

than in the case of a discharge, so that the thresholds should be proportion-
ally lower, or should be scaled to obtain uniformity in the system.

4.3. Fish risk

Figure 14 (bottom) shows the results of the ‘fish risk indicator’ for Episode
3. The indicator enters a situation of concern about the middle of the study
period, and the obtained values are growing up beyond the threshold of
“Moderate risk”. However, the detected discharge of ammonium is not sig-
nificant enough to generate a dangerous situation. From the 25th day the
risk values decrease to reach a “Normal” situation in the middle of the 29th.

As it can be observed, the ‘eutrophication indicator’ draws a gradual pro-
cess of degradation of the quality conditions in the river reaching the “Ad-
vanced eutrophication” state. Under these conditions the pH is slightly basic
which is manifested in a “Moderate risk” label for the ‘risk fish indicator’.
This represents only a risk situation, but the conditions necessary to trigger
a situation of danger to fish life exist if there is an increased concentration
of ammonium, and this could occur at any time by an urban discharge.

It has been observed in other case studies that in a situation of “Slight
eutrophication” there has been a major discharge of ammonium, but the
‘fish risk indicator’ has not reached high-risk values. The reason for this
phenomenon is that, although there is a high concentration of ammonium,
environmental conditions with normal pH do not facilitate the formation of
ammonia.

Results obtained for the ‘fish risk indicator’ suggest using the thresholds
defined in Table 3 to characterize the detected events.

The developed ‘fish risk indicator’ is a novel helpful tool for better en-
vironmental management, as it provides objective criteria for risk mitiga-
tion actions, such as forcing the dilution of nutrients to break the cycle of
eutrophication by providing flow from reservoirs, thereby avoiding possible
river wildlife mortality.

20



5. Conclusions

From the methodological point of view, algorithms based on fuzzy logic
that allow the detection and identification of episodes of urban or WWTP dis-
charge, eutrophication and fish risk have been developed. These algorithms
have been obtained from the extraction of the experience and knowledge of
experts in this field, as well for the characterization of the phenomena as
for the selection of variables to use. The presented work demonstrates the
ability of fuzzy-logic-based methods to synthesize complex information, in-
terpretable only by a few experts, and translate it into more understandable
indicators for environmental managers and the general public.

The developed expert system is based on data from measuring stations
belonging to SAICA networks. Using a few physical-chemical variables con-
tinuously recorded, the expert system is able to obtain indicators of water
quality phenomena that may be associated, with a high probability of cause-
effect relationship, to human pressure on the water environment, as would
be urban discharges or agricultural diffuse pollution. In this sense, SAICA
networks for continuous water quality measurements complement adminis-
trative official networks of manual sampling and analytical laboratory, by
providing information about episodic events (discharges) or continuous pro-
cesses (eutrophication) which can hardly be detected by discrete sampling.

The next steps to perform are the extension of the methods of calcula-
tion of indicators to more phenomena of water quality, particularly follow-up
downstream episodes, and application in other basins with different dynamic
behavior.

Acknowledgments

The SAICA network data have been provided by the Guadiana Hydro-
graphic Confederation, which has also contributed its experience and knowl-
edge of the basin needed to characterize the episodes. The work has been
developed in the context of the project ECOWATCH, led by Adasa Sistemas
S.A.U., which has received a grant (ref. 022/SGTB/2007/6.1) from “En-
vironmental projects of scientific research, technological development and
innovation” of the Spanish Ministry of Environment, Rural and Marine En-
vironment under the National Program of Environmental Science and Tech-
nology.

21



References

European Parliament, C., December 2000. Directive 2000/60/ec of the eu-
ropean parliament and of the council of 23 october 2000 establishing a
framework for community action in the field of water policy. Official Jour-
nal L 327, 1–73.

Hatzikos, E., Hätönen, J., Bassiliades, N., Vlahavas, I., Fournou, E., 2009.
Applying adaptive prediction to sea-water quality measurements. Expert
Systems with Applications 36 (3, Part 2), 6773 – 6779.

Lermontov, A., Yokoyama, L., Lermontov, M., Machado, M. A. S., 2009.
River quality analysis using fuzzy water quality index: Ribeira do iguape
river watershed, brazil. Ecological Indicators 9 (6), 1188 – 1197.

Nasiri, F., Maqsood, I., Huang, G., Fuller, N., 2007. Water quality index:
A fuzzy river-pollution decision support expert system. Journal of Water
Resources Planning and Management 133 (2), 95–105.

Serramià, A., 2005. Information systems and water quality in Spain. Inge-
nieŕıa Qúımica 420, 155–158, in Spanish.

Smith, V. H., Tilman, G. D., Nekola, J. C., 1999. Eutrophication: impacts
of excess nutrient inputs on freshwater, marine, and terrestrial ecosystems.
Environmental Pollution 100 (1-3), 179 – 196.

22