evaluation of water quality in the chillan river...
TRANSCRIPT
Environmental Monitoring and Assessment (2005) 110: 301–322DOI: 10.1007/s10661-005-8064-1 c© Springer 2005
EVALUATION OF WATER QUALITY IN THE CHILLAN RIVER(CENTRAL CHILE) USING PHYSICOCHEMICAL PARAMETERS
AND A MODIFIED WATER QUALITY INDEX
PATRICK DEBELS1,2,∗, RICARDO FIGUEROA1, ROBERTO URRUTIA1,RICARDO BARRA1 and XAVIER NIELL3
1Center for Environmental Sciences EULA-CHILE, University of Concepcion, Chile; 2VVOB,Flemish Cooperation Agency, Handelsstraat 31, Brussels, Belgium; 3Department of Ecology,
Faculty of Sciences, University of Malaga, Campus Teatinos, Malaga, Spain(∗author for correspondence, e-mail: [email protected])
(Received 16 June 2004; accepted 22 December 2004)
Abstract. The Chillan River in Central Chile plays a fundamental role in local society, as a sourceof irrigation and drinking water, and as a sink for urban wastewater. In order to characterize thespatial and temporal variability of surface water quality in the watershed, a Water Quality Index(WQI) was calculated from nine physicochemical parameters, periodically measured at 18 samplingsites (January–November 2000). The results indicated a good water quality in the upper and middleparts of the watershed. Downstream of the City of Chillan, water quality conditions were criticalduring the dry season, mainly due to the effects of the urban wastewater discharge. On the basisof the results from a Principal Component Analysis (PCA), modifications were introduced into theoriginal WQI to reduce the costs associated with its implementation. WQIDIR2 and WQIDIR, whichare both based on a laboratory analysis (Chemical Oxygen Demand) and three (pH, temperature andconductivity), respectively, four field measurements (pH, temperature, conductivity and DissolvedOxygen), adequately reproduce the most important spatial and temporal variations observed withthe original index. They are proposed as useful tools for monitoring global water quality trendsin this and other, similar agricultural watersheds in the Chilean Central Valley. Possibilities andlimitations for the application of the used methodology to watersheds in other parts of the world arediscussed.
Keywords: Central Chile, Chillan River, PCA, physicochemical parameters, water quality index
1. Introduction
Due to the worldwide concern that good quality freshwater may become a scarceresource in the near future, developing countries and countries with transitioneconomies have increased their interest in water quality monitoring programs duringthe past decades (Pesce and Wunderlin, 2000; Bordalo et al., 2001; Jonnalagaddaand Mhere, 2001). In Chile (South-America) different kinds of environmentalproblems, caused by a disordered economic growth and the excessive water useassociated with it, are affecting both the availability and the quality of freshwater.
The Chillan River, located in Central Chile, drains an important agricultural wa-tershed. Considerable amounts of water are extracted for use in irrigation agriculture
302 P. DEBELS ET AL
and for the production of drinking water for the City of Chillan. The river also re-ceives the city’s urban waste water, which – until recently – was discharged withoutany previous treatment. Surface water pollution in the watershed is supposed to bemainly related to the disposal of organic waste and the low dilution capacity of theriver during the dry period, when discharge rates are low.
Traditional approaches to assessing river water quality are based on the compar-ison of experimentally determined parameter values with the existing local norma-tive. In many cases, the use of this methodology allows for a proper identificationof contamination sources, and may be essential for checking legal compliance;however, it does not readily give a global vision on the spatial and temporal trendsin the overall water quality in a watershed.
Several authors have proposed the use of a Water Quality Index (WQI) as ameans to derive a numerical expression for the general quality of a surface water(Brown et al., 1970; Ott, 1978; Miller et al., 1986; Bordalo et al., 2001; Cude, 2001;Hallock, 2002). A single WQI value makes information more easily and rapidlyunderstood than a long list of numerical values for a large variety of parameters.Additionally, WQI’s also facilitate comparison between different sampling sitesand/or events. Consequently, they are considered better for transmitting informationto general audiences (Stambuck-Giljanovic, 1999). When their specific character-istics and limitations are taken into consideration (Ott, 1978; Flores, 2002; Hallock,2002; Pesce and Wunderlin, 2002), WQI’s can be very useful for the purpose ofmanagement and decision-making.
The use of a WQI was initially proposed by Horton (1965) and Brown et al.(1970). Since then, many different methods for the calculation of WQI’s have beendeveloped. In general, they all consider similar physical and chemical parametersbut differ in the way the parameter values are statistically integrated and interpreted(Zagatto et al., 1998; Stambuck-Giljanovic, 1999).
Previous studies on the use of a WQI in Chile are inexistent. The only well-documented example of intensive water quality-monitoring is the Biobıo RiverMonitoring Program: values for more than 30 parameters are obtained three timesa year and interpreted in a qualitative way to define a single water quality class forany given measuring site (Parra et al., 2004).
The results presented in this paper are based on physicochemical waterquality parameters determined in the Chillan River system during a period ofapproximately one year (January–November 2000). On the basis of these data, thewater quality conditions in the Chillan River system are analyzed. The data set isused to calculate a WQI, which reflects the spatial and temporal variation of thegeneral water quality in the study area. To see which parameters used in the WQIcalculations are correlated and which are responsible for most of the varianceobserved in the water quality data, a Correlation and a Principal ComponentAnalysis (PCA) are conducted. Modifications are then introduced into the indexcalculations, as to reduce the associated costs. Results from both the ‘standard’ and‘simplified’ versions of the WQI are compared and discussed, and the possibilities
EVALUATION OF WATER QUALITY IN THE CHILLAN RIVER 303
and limitations for the application of the methodology to other watersheds in Chileand other parts of the world are highlighted.
2. Materials and Methods
2.1. STUDY AREA AND SAMPLING SITES
The Chillan River Watershed (Figure 1) is located in the Biobıo Region, CentralChile (71◦00–72◦30 W, 36◦30–37◦00 S). It occupies a total surface area of 757 km2.Elevation ranges from 3200 m.a.s.l. in the Andes to 75 m.a.s.l. in the CentralValley. The total length of the Chillan River is 105 km. The climate in the area isMediterranean with a dry and warm summer period and a cold and humid winterperiod, both of approximately equal length. In the lower part of the watershed,the mean maximum temperature during January, the hottest month, is 28.8 ◦C. Themean minimum temperature for July, the coldest month, is 3.5 ◦C. Average annualprecipitation is 1624 mm. Most rainfall occurs between May and September. Meanannual river flow for the Chillan River is 22.9 m3 s−1. Winter discharge can risewell above 100.0 m3 s−1, while summer discharges as low as 1.0 m3 s−1 have beenrecorded. The Andean part of the watershed is mainly covered by native forests,while the land use in the Central Valley is dominated by agriculture (sugar beetand cereals). The City of Chillan (162.000 inhabitants) functions as a service anddistribution center for the agro-business. No important industrial activity takes placewithin the watershed.
To characterize water quality and its’ spatial variability along the river network,locations for 18 sampling stations were carefully selected. From these, eight siteswere defined on the main course, while the remaining 10 sites were located ontributaries (Figure 1).
The first sampling station for the Chillan River (E1) was located at the pointwhere the river leaves the Andes Mountain Range and enters the Central Valley
Figure 1. Location of the Chillan Watershed (left); location of the different sampling stations (right).
304 P. DEBELS ET AL
(76.6 km – measured along the river starting from the outlet). The river is expectedto have good water quality conditions here, as human activities upstream of thispoint are limited. Station E2 (57.1 km) is located directly downstream of the smallurban center of Pinto, while station E3 (48.0 km) reflects the water quality near theintake of the drinking water supply for the City of Chillan. Station E4 (34.7 km)and E5 (31.8 km) are upstream and south of the city centre, respectively. Station E6(26.6 km) is 3.6 km upstream of the Las Toscas Tributary, which receives the city’surban wastewater only a short distance (ca. 500 m) before it enters the ChillanRiver. Station E7 (16.8 km) is 6.2 km downstream of the confluence of the LasToscas Tributary with the Chillan River, and 1.1 km downstream of the QuilmoRiver, the most important tributary. Station E8 (3.2 km) indicates conditions in theChillan River near the outlet of the watershed.
For the tributaries in the Central Valley, sampling stations were assigned to theirmiddle reaches (T1-T2-T3-T4-T5-T8) and to a point near their confluence withthe Chillan River (T6-T7-T9-T10). The station on the Las Toscas River (T10) islocated just below the waste water discharge. The tributaries in the Andean partwere not considered in this study, as their water quality conditions were supposedto be reflected by the water quality in the main river at station E1.
Temporal variability of water quality was evaluated by realizing sampling cam-paigns at a two-months interval, starting in January 2000 and ending in November2000. March, at the end of the Austral summer, typically represents the lowestdischarge rates, while frequent rainfall (autumn and winter) and snow-melt (spring)lead to much higher discharges between April–May and September–October.
2.2. SAMPLE COLLECTION AND ANALYSIS
All physicochemical parameters were sampled and determined according to stan-dard methods (APHA-AWWA-WPCF, 1995). Table I gives an overview of theparameters determined during this study. Temperature, pH and conductivity weremeasured in the field using a thermometer with a precision of 0.1 ◦C, a calibratedportable pH-meter “Schott Gerate”, and a digital conductivity meter “Cole Palmer”.
For the calculation of the WQI, the following equation was used (Martınez deBascaran, 1979):
WQI =∑n
i=1 Ci ∗ Pi∑n
i=1 Pi
where n represents the total number of parameters, Ci is the value assigned to param-eter i after normalization, and Pi is the weight of the parameter (an indicator of its’relative importance for aquatic life/human water use). In the above equation, WQIis calculated as the weighted sum of the different subindex scores. Several alterna-tive methods have been proposed to aggregate the individual parameter scores intoa final index value. The (dis)advantages of each method, such as e.g. the eclipsing
EVALUATION OF WATER QUALITY IN THE CHILLAN RIVER 305TA
BL
EI
Mea
nan
nual
valu
esfo
rph
ysic
alan
dch
emic
alw
ater
qual
itypa
ram
eter
s(J
anua
ry–N
ovem
ber
2000
)
Var
iabl
eSt
atio
nE
1E
2E
3E
4E
5E
6∗E
7E
8T
1T
2T
3T
4T
5T
6T
7T
8T
9T
10
pH(n
=6)
Mea
n7.
67.
87.
98.
17.
68.
57.
27.
57.
07.
06.
86.
46.
97.
07.
27.
27.
27.
3S.
D.
0.5
0.8
0.9
1.0
0.4
0.2
0.3
0.9
0.5
0.4
0.4
0.4
0.5
0.6
0.6
0.4
0.5
0.3
Tem
pera
ture
(n=
6)M
ean
11.2
14.5
14.1
15.6
14.9
20.7
16.2
17.5
13.6
13.5
12.6
13.3
14.2
14.8
15.3
12.2
13.1
18.1
(◦ C)
S.D
.3.
45.
76.
77.
15.
70.
56.
67.
43.
23.
23.
33.
43.
34.
34.
92.
42.
95.
0C
ondu
ctiv
ity(n
=6)
Mea
n67
.770
.669
.179
.670
.892
.714
5.0
130.
326
.729
.427
.331
.436
.845
.577
.442
.048
.537
6.0
(µS
cm−1
)S.
D.
27.4
27.6
29.5
31.2
24.6
16.5
130.
080
.07.
78.
44.
88.
36.
622
.520
.518
.918
.916
8.2
DO
(n=
6)M
ean
10.2
10.4
10.5
10.7
10.5
8.7
8.6
8.0
10.0
9.5
9.6
8.6
9.3
9.1
9.9
10.1
9.3
2.8
(mg
L−1
)S.
D.
0.8
0.7
0.9
1.0
1.5
0.4
2.4
2.4
1.1
0.9
1.0
1.5
0.7
1.8
0.6
0.8
1.1
1.8
DO
(%sa
t)(n
=6)
Mea
n92
.810
1.1
101.
110
0.9
103.
296
.985
.678
.796
.090
.389
.589
.190
.581
.393
.687
.998
.729
.0S.
D.
2.2
6.2
7.9
7.1
13.8
5.7
20.5
20.4
7.8
3.8
5.3
15.3
2.6
9.8
3.6
5.7
8.3
16.8
BO
D5
(n=
6)M
ean
1.5
1.1
0.9
2.3
1.7
3.3
4.3
4.5
1.3
1.1
1.1
0.9
1.0
1.1
1.3
1.1
1.2
64.3
(mg
L−1
)S.
D.
0.5
0.1
0.4
1.4
1.2
3.2
4.9
3.9
0.4
0.2
0.3
0.2
0.4
0.2
0.5
0.4
0.4
54.4
CO
D(n
=6)
Mea
n1.
71.
71.
23.
53.
77.
014
.813
.23.
85.
55.
24.
64.
37.
55.
23.
05.
218
9.4
(mg
L−1
)S.
D.
0.8
0.8
0.7
1.6
2.5
2.8
15.5
7.7
2.5
4.0
1.6
0.9
2.1
8.7
1.3
1.9
2.1
105.
1A
mm
onia
(n=
6)M
ean
0.01
20.
012
0.01
10.
025
0.01
90.
352
3.17
20.
793
0.01
30.
013
0.01
50.
037
0.05
10.
019
0.01
60.
014
0.06
118
.579
(mg
NL
−1)
S.D
.0.
003
0.00
40.
002
0.02
20.
006
0.48
75.
120
1.05
80.
007
0.00
40.
004
0.03
20.
078
0.00
90.
006
0.00
40.
103
7.40
4N
itrite
s(n
=6)
Mea
n0.
0003
0.00
030.
0005
0.00
260.
0020
0.02
250.
0256
0.11
470.
0009
0.00
120.
0011
0.00
080.
0016
0.00
290.
0035
0.00
130.
0017
0.07
65(m
gN
L−1
)S.
D.
0.00
020.
0002
0.00
020.
0028
0.00
090.
0308
0.02
440.
1075
0.00
030.
0007
0.00
070.
0002
0.00
080.
0045
0.00
180.
0006
0.00
080.
1179
Nitr
ates
(n=
6)M
ean
0.00
60.
048
0.06
10.
166
0.12
50.
110
0.18
50.
683
0.19
90.
237
0.19
60.
280
0.36
70.
131
0.67
40.
237
0.22
20.
644
(mg
NL
−1)
S.D
.0.
006
0.01
90.
047
0.04
80.
121
0.15
10.
167
0.46
20.
107
0.13
00.
143
0.21
00.
183
0.12
10.
536
0.19
10.
177
1.52
8O
rgan
icN
(n=
6)M
ean
0.04
30.
115
0.12
20.
305
0.26
30.
815
3.76
51.
997
0.36
50.
410
0.33
20.
528
0.58
20.
357
0.93
00.
464
0.42
326
.390
(mg
NL
−1)
S.D
.0.
023
0.04
70.
026
0.06
90.
065
0.89
85.
248
1.14
00.
085
0.12
40.
148
0.23
90.
173
0.25
20.
435
0.20
60.
215
9.32
0O
rtho
phos
phat
e(n
=5)
Mea
n0.
013
0.00
70.
006
0.00
80.
009
0.08
80.
603
0.36
40.
003
0.00
30.
002
0.00
30.
003
0.01
30.
008
0.00
40.
005
2.76
5(m
gP
L−1
)S.
D.
0.00
50.
003
0.00
30.
004
0.00
50.
116
1.01
70.
404
0.00
20.
001
0.00
00.
003
0.00
30.
011
0.00
30.
003
0.00
31.
179
Org
anic
P(n
=6)
Mea
n0.
028
0.01
30.
012
0.01
80.
020
0.10
30.
632
0.39
90.
009
0.01
20.
011
0.00
90.
014
0.03
10.
026
0.01
40.
019
3.96
7(m
gP
L−1
)S.
D.
0.02
80.
004
0.00
40.
004
0.00
20.
117
1.07
40.
371
0.00
20.
003
0.00
50.
002
0.00
70.
021
0.00
60.
006
0.00
71.
622
Har
dnes
s(m
g(n
=3)
Mea
n17
.418
.219
.321
.320
.326
.621
.923
.88.
89.
511
.410
.912
.612
.926
.313
.214
.480
.6C
aCO
3L
−1)
S.D
.4.
74.
43.
93.
83.
7–
2.4
0.9
0.8
1.6
2.1
1.3
2.9
3.6
3.4
1.7
1.8
6.2
Oils
and
fats
(n=
1)<
0.10
<0.
10<
0.10
<0.
10<
0.10
<0.
10<
0.10
<0.
10<
0.10
<0.
10<
0.10
<0.
10<
0.10
<0.
10<
0.10
<0.
10<
0.10
4.1
(mg
L−1
)C
alci
um(n
=1)
6.4
6.8
7.0
7.6
7.4
–9.
69.
83.
23.
63.
62.
84.
22.
89.
25.
05.
020
.6(m
gL
−1)
Tur
bidi
ty(n
=1)
1.6
1.6
2.5
9.6
6.6
–6.
15.
87.
15.
64.
26.
83.
76.
87.
014
.118
.274
.5(N
TU
)
n=
num
ber
ofca
mpa
igns
;S.D
.:St
anda
rdde
viat
ion.
∗ Sta
tion
E6
was
sam
pled
only
twic
edu
ring
the
entir
est
udy
peri
od.
306 P. DEBELS ET AL
or over-emphasizing of a single bad parameter value, have been thoroughly dis-cussed in literature (Hallock, 2002), and have to be taken into consideration wheninterpreting the final results.
Table II gives the different parameters that were used in the evaluation process,as well as their relative weights and the normalization factors (parameter ratingcurves). The values for the weight and normalization factors were adapted to localconditions, using expert judgment and a variety of sources from literature (e.g., Ott,1978; Conesa, 1997; Pesce and Wunderlin, 2000; Cude, 2001; Ward, 2001). Dueto the lack of available field data, the WQI proposed in this work does not take intoaccount suspended solids, microbiological contamination nor toxic compounds;only parameters for which a complete data set was available for the study area wereconsidered.
WQI values express freshwater quality as a percentage of an optimum situation,with 100% being the best possible quality. For each station, one annual and threeseasonal WQI values were determined. To calculate the summer WQI at a givenstation, average parameter values were calculated from the data obtained during theJanuary and March field campaigns. In the same way, a winter and a spring WQIwere obtained by using the data from May–July, and from September–November,respectively. Values from all six campaigns were averaged for calculating the annualWQI. The obtained values can be interpreted both in a quantitative or a qualitativeway. In the last case, values between 100% and 80% are considered to correspondto ‘very good general water quality’ conditions, while subsequent 20% intervalsare classified as ‘reasonably good’, ‘deteriorated’, ‘bad’ and ‘very bad’.
Both Biological (BOD5) and Chemical Oxygen Demand (COD) were measuredduring the sampling campaigns. Considering the fact that COD can be determinedin a more accurate way and at a lower cost than BOD5, the effect of the replacementof BOD5 by COD in the WQI calculations was evaluated. A Principal ComponentAnalysis (PCA) was then conducted on the log-transformed measurement data,using all WQI parameters. The results from this analysis were used to proposemodifications to the WQI, which help to reduce the amount of parameters requiredfor its calculation, as well as the analytical costs associated with it. The signifi-cance of the differences observed between the different WQIs was evaluated usingWilcoxon’s paired sample test. On the basis of the previous results, the performanceand generalized applicability of the modified WQI’s are discussed.
3. Results and Discussion
Mean annual values and standard deviations (S.D.) for the different water qualityparameters are given in Table I. The results from the seasonal and annual standardWQI calculations are shown in Figure 2, for the Chillan River (E1–E8) and itstributaries (T1–T10), respectively. Station E6 is not shown in the graphics as it wassampled only twice.
EVALUATION OF WATER QUALITY IN THE CHILLAN RIVER 307
TAB
LE
IIV
aria
bles
used
inth
eW
ater
Qua
lity
Inde
xca
lcul
atio
ns,s
core
sof
norm
aliz
atio
nan
dre
lativ
ew
eigh
ts
Rel
ativ
eN
orm
aliz
atio
nFa
ctor
(Ci)
wei
ght
Var
iabl
e(P
i)10
090
8070
6050
4030
2010
0
DO
(mg
L−1
)0.
18≥7
.5>
7.0
>6.
5>
6.0
>5.
0>
4.0
>3.
5>
3.0
>2.
0≥1
.0<
1.0
BO
D5
(mg
L−1
)0.
17<
0.5
<2.
0<
3.0
<4.
0<
5.0
<6.
0<
8.0
<10
.0<
12.0
≤15.
0>
15.0
Am
mon
ia(m
gN
L−1
)0.
13<
0.01
0<
0.02
5<
0.05
0<
0.10
0<
0.20
0<
0.30
0<
0.50
0<
0.75
0<
1.00
0≤1
.250
>1.
250
Nitr
ates
(mg
NO
3−
L−1
)a0.
07<
0.5
<2.
0<
4.0
<6.
0<
8.0
<10
.0<
15.0
<20
.0<
40.0
≤70.
0>
70.0
Nitr
ites
(mg
NO
2−
L−1
)a0.
07<
0.00
5<
0.00
8<
0.01
0<
0.04
0<
0.07
5<
0.10
0<
0.15
0<
0.20
0<
0.25
0≤0
.500
>0.
500
Ort
hoph
osph
ate
(mg
PL
−1)
0.12
<0.
025
<0.
050
<0.
100
<0.
200
<0.
300
<0.
500
<0.
750
<1.
000
<1.
500
≤2.0
00>
2.00
0
Tem
pera
ture
(◦ C)
0.10
<20
.0<
21.0
<22
.0<
24.0
<26
.0<
28.0
<30
.0<
32.0
<36
.0≤4
0.0
>40
.0
pH0.
107.
06.
9–7.
56.
7–7.
86.
5–8.
36.
2–8.
75.
8–9.
05.
5–9.
55.
0–10
.04.
5–10
.54.
0–11
.5<
4.0;
>11
.5
Con
duct
ivity
(µS
cm−1
)0.
06<
600
<70
0<
850
<10
00<
1250
<15
00<
2000
<25
00<
3000
≤350
0>
3500
CO
D(m
gL
−1)b
0.17
<1.
2<
5.0
<7.
5<
10.0
<12
.5<
15.0
<20
.0<
25.0
<30
.0≤4
0.0
>40
.0
a Mea
sure
dni
trat
ean
dni
trite
valu
esar
eco
nver
ted
from
mg
NL
−1to
mg
NO
− xL
−1pr
ior
tous
ing
the
cite
dno
rmal
izat
ion
fact
ors.
bC
OD
ispr
opos
edas
anal
tern
ativ
eto
the
use
ofB
OD
5fo
rthe
calc
ulat
ion
ofth
eW
QI,
and
has
norm
aliz
atio
nfa
ctor
sth
atar
eba
sed
ona
regr
essi
onan
alys
isbe
twee
nbo
thpa
ram
eter
s.
308 P. DEBELS ET AL
Figure 2. Results of the seasonal and annual Water Quality Index calculations (WQIC O D) for theChillan River and its tributaries.
3.1. WQI CALCULATIONS
WQI values were calculated using both a standard methodology based on BOD5
and an alternative method where BOD5 is replaced by COD. At some stations, thecalculations from both methodologies resulted in slightly different values, but thesewere statistically not significant (mean diff = 0.37; S.D. = 1.49; p < 0.05). Forthis particular study, the methodology based on COD was proposed for further use.
Most obtained WQI values are higher than 80% and in many cases they are closeto the optimum of 100%. Summer values, however, are generally lower (p < 0.05),especially in the middle and lower part of the watershed. Downstream of stationE3, water quality in the main course begins to deteriorate significantly. The effectsof the presence of the City of Chillan become specifically clear at E7–E8; at E8 theminimum WQI for the main river is reached (∼40%) during the summer period.According to the index, most tributaries have a very good general water quality
EVALUATION OF WATER QUALITY IN THE CHILLAN RIVER 309
throughout the year. Only at T10, downstream from the waste water discharge,water quality is poor year-round. This is the site where the lowest index values,around 35%, were obtained. No statistically significant seasonal differences in indexvalues could be detected for the stations located on the tributaries.
3.2. PHYSICOCHEMICAL WATER QUALITY
In order to get a better view on the causes of the deteriorated water quality inthe downstream part of the watershed, selected results from the determination ofphysicochemical parameters are discussed below.
Figure 3 shows mean annual levels of inorganic N-compounds and orthophos-phates. Levels remain low down to station E5. At E7, ammonia concentrationsincrease up to 167 times the value of E5. After E7, they decrease rapidly. Nitrite in-creases 57 times between E5 and E8. Nitrate concentrations at E8 are almost 6 timeshigher than at E5. The behavior presented by these components is indeed very typi-cal for reaches located immediately downstream of an urban waste water discharge:the input of untreated sewage initially gives rise to high levels of organic nitrogenand ammonia (T10). Downstream, organic nitrogen is further converted to ammonia– reflected in the high concentrations at E7 – which then gradually transforms intonitrite and finally nitrate through a nitrification process, during which oxygen isconsumed (Chapra, 1997). Ammonia is also eliminated through uptake by macro-phytes that proliferate in this area. This process is more efficient and up to threetimes faster than nitrate uptake (Reddy and Tucker, 1983; Sastre et al., 1998; Stanleyand Hobbie, 1981), so both nitrite and nitrate concentrations can still rise betweenE7 and E8, due to nitrification. Orthophosphate levels show more or less the sametrend as ammonia, reaching a maximum concentration at E7, followed by a substan-tial drop at E8. The standard deviations (S.D.) given in Table I show that the generaltrend in water quality, as described before, is valid throughout the year in the upperpart of the watershed, but seasonal variations do occur at the downstream stations.
Both phosphorus and inorganic nitrogen play an important role in the eutrophica-tion process of the receiving surface waters (Soulsby et al., 2001). Orthophosphatescan quickly be absorbed by plants and generally have a greater influence on eutroph-ication than nitrogen (Margalef, 1983; Elser et al., 1999; Sharpley et al., 2001).Ammonia, on the other hand, is very toxic to fish when present in its un-ionizedform (pH and temperature dependent), even at very low concentrations.
Between E7 and E8, the development of mantles of macrophytes may covermore than 50% of the cross-sectional area during summer. The high concentrationsof ammonia and orthophosphates stimulate plant growth here, with photosynthesisand respiration playing an important role in the self-purification process of the river(Jonnalagadda and Mhere, 2001).
Station E6, upstream of the urban waste water discharge, already shows ele-vated levels of orthophosphates, ammonia and nitrites, as compared to E5 (TableI). This seems to indicate the existence of one or several smaller, non-identified
310 P. DEBELS ET AL
Figure 3. Mean annual inorganic nitrogen and phosphorus concentrations in the Chillan River andits tributaries.
organic contamination sources between E5 and E6, probably due to the clandestineevacuation of urban waste waters. Periphyton and macrophytes – although moreabundant at E7 – were already present near E6. It should be stated however thatE6 was only sampled 2 times during the entire study period, so care should betaken when comparing mean parameter values with those from other stations. Onthe basis of the data obtained during these campaigns, its inclusion as a permanentstation in future monitoring schemes is recommended.
Although nitrate levels are generally higher than in the main stream, mean an-nual levels of inorganic nitrogen and orthophosphates in the tributaries remain low
EVALUATION OF WATER QUALITY IN THE CHILLAN RIVER 311
for all stations, except T10 where a drastic rise in ammonia and orthophosphatescan be observed, caused by the discharge of urban waste water. The effect is tobe seen in the Chillan River (Figure 3), which receives the contribution of the LasToscas Tributary between E6 and E7. The higher value for nitrates observed in thelower reach of the Cadacada Tributary (T7) seems to be reflected downstream inthe Chillan River, at E4. The same tendency is not – or only to a lesser extend – ob-served for orthophosphate and ammonia. This may indicate the existence of diffusecontamination from agriculture in this sector, as phosphate sources would be fixedto the soil matrix and ammonia would be quickly eliminated through volatilization,plant uptake and nitrification. Water quality at E7 seems to be principally affectedby the input of urban waste water at T10, whereas the contribution of the QuilmoTributary (T9) causes no major effect.
Variations of DO, BOD5 and COD along the river network are illustrated inFigure 4. Again, all values indicate good water quality conditions in the upperreaches. Levels of organic contamination become more important downstream fromstation E3: at E7 the effect of the discharge of urban sewage is very obvious.Although other parameter values already increased at E6, DO levels here are stillnear saturation. Probably, at this station, during the day oxygen consumption bybiodegradation processes is compensated for by reaeration and photosynthesis.The somewhat lower values of daytime DO at E7 and E8 – compared to saturationlevels – in spite of high photosynthetic activity in this sector can be explained by theimportance of biodegradation processes (including nitrification), and the presenceof some larger pools in this reach. Reaeration in the Chillan River is generally high(pool-riffle structure), and DO levels in the main stream never reach critical values,not even in the most contaminated part. However, it is important to mention thatall samples were taken during daytime, and no evaluation of diurnal fluctuations ofDO levels has been made so far.
Spatial and temporal variation of pH, temperature and conductivity can be de-duced from Figure 5 and from the values for standard deviation in Table I. The meanannual values for pH range from 6.4 to 8.5, which is within the limits of the naturalvalues that support the region’s aquatic life (Hellawell, 1986; Allan, 1997; Nagelset al., 2002)). For the main stream, the lowest value is observed downstream ofthe waste water discharge, and is probably due to decomposition of organic matter.Mean annual water temperature varies between 11.2 ◦C (in the upper part of the wa-tershed) and 17.5 ◦C (in the lower part). Temporal variability is more important at thedownstream stations than at E1, where the river recently leaves the Andes MountainRange. From here on, the river widens, becomes slower, and shading by vegetationis less important. Shading however may still be important on the smaller tributarieslocated in the Central Plain, where seasonal variations in water temperature tendto be smaller. The downstream temperature gradient in the main river may also beinfluenced by the sampling schemes, which were typically initiated in the upperparts of the watershed. Conductivity is low where the discharge is formed by meltwater coming from the snow-capped peaks in the Andes. The poorly mineralized
312 P. DEBELS ET AL
Figure 4. Mean annual Dissolved Oxygen (DO), 5-day Biochemical Oxygen Demand (BOD5) andChemical Oxygen Demand (COD) for the Chillan River and its tributaries.
waters become increasingly ionized towards the lower reaches, but the values re-main beneath 150 µS · cm−1. A more pronounced difference can be observed inthe tributaries, where normal conductivity ranges from 26.5 to 77.2 µS · cm−1 butwhere a local maximum of 376.0 µS · cm−1 occurs due to the discharge of urbanwaste water.
Some water quality variables measured during the first campaigns were notconsidered in later samplings. These include turbidity (n = 1; 1.6–18.2 NTU),Calcium (n = 1; 2.8–9.6 mgL−1) and hardness (n = 3; 8.72–26.3 mgL−1 CaCO3).
EVALUATION OF WATER QUALITY IN THE CHILLAN RIVER 313
Figure 5. Mean annual temperature, pH and conductivity for the Chillan River and its tributaries
The measured values coincide with the typical water characteristics of CentralChile, which means very soft waters of nivo-pluvial origin that run on graniterocks. Fats and oils were beneath the detection limit (n = 1; <0.10 mgL−1) at allstations, with only one exception (T10). No important indices of contamination byorganochlorine pesticides could be detected.
3.3. PCA ANALYSIS AND PROPOSED WQI MODIFICATIONS
In order to see which parameters used in the WQI calculations are correlated andwhich are responsible for most of the variance observed in the water quality data, a
314 P. DEBELS ET AL
TABLE IIIEigenvalues and factor loadings from the Principal Component Analysis (boldvalues >0.80)
Factor 1 Factor 2 Factor 3
PCA FactorEigenvalues 5.33 1.85 1.40
Percentage of variance 53.28 18.52 14.01
Accumulative % 53.28 71.18 85.82
Factor loadings (varimax normalized)
NH4 0.98 0.04 0.06
BOD5 0.90 −0.02 0.00
Conductivity 0.88 0.15 0.25
COD 0.98 −0.05 −0.01
DO −0.80 −0.21 −0.23
Nitrate 0.00 0.90 −0.25
Nitrite 0.22 0.91 0.14
pH 0.00 −0.18 0.87
Orthophosphate 0.86 0.39 0.10
Temperature 0.23 0.10 0.87
Total proportion 4.97 1.79 1.73
Principal Component Analysis (PCA) was conducted. Considering the skewness ofsome parameters’ sample data distributions – due to the presence of outliers (pointsources) in the original data set – the PCA analysis was performed using the log-transformed data. The results of the calculations are given in Table III and Figure 6.From the figure, it can be seen that by using the first 2 components, the measurementsfrom the 6 campaigns realized at station T10 (T10-1 to T10-6), and from the summercampaigns at E7 and E8 become dispersed over the bi-dimensional space, separatedfrom the cluster formed by the rest of the data. A similar result is obtained when us-ing Factor 1 and 3, but the ‘clustered’ data become more stretched along the Y-axis.Factor 1 explains 53.3% of the total variance in the data set. The spreading along thehorizontal axis is mainly due to the variance in the parameter values for ammonia,orthophosphate, conductivity, COD, DO and BOD5 and reflects the changes dueto contaminant input from point sources, combined with seasonal effects. Factor 2,which is heavily influenced by nitrite and nitrate levels, explains an additional 18.5%of the variance, while temperature and pH are the main parameters responsible forthe spreading observed along Axis 3 (14.0%). The distribution of the measurementsalong Axis 3 is quite continuous, and indicates the predominance of gradual changesin factor values due to the existence of e.g. a downstream temperature gradient(Figure 5). The three factors together explain 85.8% of the variance in the originaldata set.
EVALUATION OF WATER QUALITY IN THE CHILLAN RIVER 315
Figure 6. Distribution of the sampling stations over a bidimensional space, using PCA Factor 1, 2and 3; the first digits in the used codification correspond to the sampling station (1–10), while the lastdigit refers to the field campaign on which the data are based (chronological ranking, from 1 to 6).
The values for COD, BOD5, ammonia and orthophosphate in the data set arehighly correlated (Table IV). Considering the additional fact that classificationintervals are used (Table II) before the WQI is calculated, it seems well worthinvestigating the impact on the results of the WQI calculations caused by the elim-ination of one or several of these parameters. As can be seen from Table III, thepreviously mentioned parameters all have similar loading values in PCA Factor 1.
316 P. DEBELS ET AL
TABLE IVPearson’s coefficient of correlation r for the different water quality parameters (analysis based onthe log-transformed data set)
Parameter pH Temp Cond DO BOD COD Ammonia Nitrates Nitrites OrthoP
pH – 0.54 0.55 −0.06 0.15 −0.11 0.00 −0.45 0.08 0.23
Temp – – 0.47 −0.42 0.29 0.33 0.29 −0.22 0.44 0.39
Cond – – – −0.62 0.68 0.55 0.68 −0.30 0.59 0.81
DO – – – – −0.70 −0.74 −0.76 0.23 −0.51 −0.72
BOD – – – – – 0.75 0.76 −0.06 0.59 0.80
COD – – – – – – 0.77 0.10 0.68 0.68
Ammonia – – – – – – – 0.00 0.73 0.85
Nitrates – – – – – – – – 0.33 −0.11
Nitrites – – – – – – – – – 0.72
If the objective is to reduce the number of parameters (and associated costs) thathave to be measured in the field, a first proposal (WQIRE D) can be made based onthe elimination of ammonia and orthophosphate. In the case of the Chillan water-shed, due to their correlation with COD, their relative weights can be added to theweight factor for COD. Inorganic p-levels may control eutrophication more thaninorganic N, elevated nitrogen levels were for long thought to represent an addi-tional environmental hazard, due to the potential risks for human health associatedwith the presence of nitrates and nitrites in drinking water (Addiscott et al., 1991;Emsley, 1994). Recent controversy on this topic (Addiscott, 1999) has shifted theemphasis concerned with nitrate and nitrite back to the widely accepted eutrophica-tion problem, and to the toxic effects of nitrite on aquatic life (Hatch et al., 2002).At this point, nitrate and nitrite are maintained as parameters in the calculationof WQIRED. The parameters that still have to be determined experimentally arethus: COD, nitrate, nitrite, DO, conductivity, temperature and pH. From these, thelast 3(4) can easily be determined in the field, reducing the amount of requiredlaboratory analyses from 5(6) to 3(4).
Results from WQIRED (not shown) are statistically not different from theoriginal calculations (r2 = 0.97; slp = 0.99; mean diff = −1.0; S.D. = 3.0;p < 0.05), except for the spring season data set. Indeed, at most stations, theindividual parameter rating curves for COD, ammonia and orthophosphates allyielded very similar scores.
A further simplification of the methodology (WQIDIR) consists in a laboratoryanalysis of a single parameter, COD, plus the direct measurement of conductiv-ity, temperature, pH and DO in the field, while nitrate and nitrite are eliminatedfrom the calculations. Considering the results from the PCA, it can be seen thatunder these conditions, most of the variance observed along Axis 1 and 3 is stilltaken into account. When the stations affected by point sources (City of Chillan)
EVALUATION OF WATER QUALITY IN THE CHILLAN RIVER 317
are excluded from the PCA analysis, temperature, conductivity and pH becomethe parameters that explain most of the variance observed along Axis 1 (data notshown). A considerable part of the longitudinal variance of these parameters maybe due to natural factors. Temperature and pH remain however very important,as human-induced changes in their natural spatial-temporal patterns may severelyaffect aquatic ecosystem health. Conductivity, on the other hand, is a sensitiveparameter easily measured in the field that throws a light on the presence of dis-solved ions beyond natural background levels. These are all important aspects thatshould be considered when establishing relative weights and class limits (scores ofnormalization, Table II) for any given river system where the WQI will be applied.
The results from the calculations of WQIDIR are shown in Figure 7. Numerically,the mean values of the seasonal and annual WQIDIR calculations are slightly inferiorto those obtained with the original index (slope = 0.97; r2 = 0.94; mean diff =−2.5; S.D. = 4.2; p < 0.05). In a qualitative analysis where discrete water qualityclasses are used per 20% index interval, the results from both the original methodand from WQIDIR would lead to mostly the same conclusions: in three occasions,
Figure 7. Results of the seasonal and annual calculations (WQIDIR; with Dissolved Oxygen).
318 P. DEBELS ET AL
the new method would clearly lead to a classification into an immediately superioror inferior water quality class, i.e. at station E8 (from ‘reasonably good’ to ‘verygood’) and station E4 (from ‘very good’ to ‘reasonably good’), both during thespring season, and at station T10 during winter (from ‘deteriorated’ to ‘bad’).
3.4. USE OF DO AS A PARAMETER IN THE WQI
In a study on the water quality of the Suquıa River, Cordoba (Argentina), Pesce andWunderlin (2000) also proposed a modified index, WQImin, calculated from onlythree parameter values: conductivity, turbidity and DO. WQImin values presentedthe same temporal and spatial tendencies as those obtained with their extendedindex, WQIobj, based on 20 parameters.
In the present work, a moderately strong (negative) correlation could be observedbetween DO and COD, ammonia and orthophosphate (r ∼ −0.73). The importanceof DO as a water quality indicator has to be stressed, as it is in deed a key factor foraquatic life. However, very often the effects of diurnal or temperature-induced DOvariations are not taken into account. Samples taken at equally ‘contaminated’ sitesbut at different moments of the day, or samples taken before, within, or just after apool, can lead to completely different diagnoses of local water quality. In nutrient-rich pools, very high DO values at the end of the day may at least be partiallyinfluenced by photosynthetic activity. During the night and early morning, severelyreduced oxygen levels – or even anoxia – may produce, leading to mortality of manyaquatic species (Helawell, 1986). These adverse conditions may not be reflected bythe WQI, as measured afternoon DO would (partially) compensate for the elevatednutrient concentrations, especially when a high relative weight is assigned to thisparameter.
In order to overcome this problem, some authors consider both the actual DOconcentrations as well as the percentage of saturation (%sat); DO levels that indicateoversaturation get a lower score (Cude, 2001). However, in many cases a possiblemisinterpretation of DO values close to saturation can not be excluded, unless thesampling scheme adheres to a strict time schedule (e.g. all DO measurements inthe morning if photosynthesis/respiration is thought to be important).
Independent of the fact whether organic contamination leads to oxygen de-pletion or not, it does contribute to the eutrophication of the ecosystem, whichultimately may result in an alternation of species composition, and a reduction ofthe native biodiversity. For this reason, in many river systems elevated levels oforganic contamination should be considered as adverse conditions, and this shouldbe adequately reflected by the WQI.
The reaeration capacity of Chilean rivers is generally high. In the present study,at most sampling sites a score of 100 was obtained for DO, after normalizationof the measured values. Many of the observed DO values are close to saturation(Table I), and in some cases indications exist of DO values being influenced byphotosynthetic activity. As field campaigns generally started in the upper part of
EVALUATION OF WATER QUALITY IN THE CHILLAN RIVER 319
the watershed and ended late afternoon in the contaminated part near the outlet, theinterpretation of medium and high DO values in this area is not straightforward.
Elimination of DO as a parameter is expected to lead to lower index values:when applied to the original WQI, a statistically significant drop is observed forall seasons, although the magnitude of this drop is rather limited (slope = 0.97;r2 = 0.98; mean diff = −2.7; S.D. = 2.5; p < 0.05) and it would lead to a differentqualitative interpretation only at a few sites (E7 and E8 during summer, from‘deteriorated’ to ‘bad’, and at T6 during spring, from ‘very good’ to ‘reasonablygood’). This shows that the initial relative weight factor of 18% for DO in the WQIcalculations may still be justified, if adaptive measures are taken during samplingthat consider possible diurnal/local variations of both DO and temperature.
Elimination of DO from WQIDIR (the remaining parameters in WQIDIR2 are:COD, conductivity, temperature and pH) leads to a similar significant reductionof the mean index values (p < 0.05), but a slightly major temporal and spatialdifferentiation is achieved (Figure 8).
Figure 8. Results of the seasonal and annual Water Quality Index calculations (WQIDIR2; withoutDissolved Oxygen).
320 P. DEBELS ET AL
4. Conclusions
• An analysis of the water quality in the Chillan Watershed by means of a WQIshowed a good general water quality in most of the watershed, throughout theyear. Severely deteriorated conditions were detected during summer in stationsdownstream of the urban waste water discharge. The most affected parameterswere COD, ammonia, nitrate and orthophosphate. Observed DO levels wereseldom critical.
• The origin of the contamination and the correlation observed between parame-ters suggest that for routine sampling only a reduced number of parameters maybe analyzed. Based on the results from a PCA, several alternative WQIs wereevaluated. The proposed modifications all led to similar results and the majorspatial and temporal trends remained well represented. In the case of the Chillanwatershed, for routine monitoring of trends in global river water quality, WQIDIR
(based on COD, DO, pH, temperature and conductivity) and WQIDIR2 (withoutDO) can be used to replace the original index, unless drastic changes in landuse and/or human activities would lead to completely different contaminant loadcompositions. WQIDIR and WQIDIR2 may also serve as a tool to monitor trends inother, similar agricultural watersheds in Chile’s Central Valley. A periodic com-parison of the results with those from the extended WQI remains recommended,in order to verify the validity of the assumptions used.
• The elimination of parameters from extended versions of WQIs based on a Cor-relation and PCA analysis is a technique that can be useful in many parts of theworld, especially in those countries where resources for operational water qualitymodeling are rather scarce. Whenever possible, however, attention should be puton obtaining an extensive basic water quality data set, which includes a widevariety of water quality parameters and which covers a broad range of waterquality conditions.
• The extended index used in this work is based on water quality informationavailable for the studied watershed. Inclusion of additional parameters such asturbidity and especially microbiological data is recommended, whenever possi-ble. Local background conditions should be taken into account when establishingparameter rating curves for a WQI. In this sense, the applicability of a specificWQI may be limited to the aquatic ecoregion/watershed for which it has beendeveloped. WQI’s should not be unrestrictedly used without consideration oftheir characteristics and limitations.
Acknowledgments
This work is the result of a cooperation project between the Aquatic Systems Unitof the EULA-CHILE Center, University of Concepcion, Chile, the VVOB Agency,Flanders, Belgium and the Department of Ecology, University of Malaga, Spain.
EVALUATION OF WATER QUALITY IN THE CHILLAN RIVER 321
Field work was developed in the framework of the project “Development of aMethodology for the Evaluation and Mitigation of the Contamination of Water andSoils: Application to the Chillan Watershed”, which was partially financed by theChilean Agriculture and Livestock Service (SAG, Project N◦ VIII 4-36-0199). Theauthors wish to thank the two anonymous reviewers for their helpful comments andsuggestions.
References
Addiscott, T. M.: 1999, Nitrate and Health, in: W. S. Wilson, A. S. Ball and R. H. Hinton (eds), Man-aging Risks of Nitrates to Humans and the Environment, Royal Society of Chemistry, Cambridge,pp. 247–249.
Addiscott, T. M., Whitmore, A. P. and Powlson, D. S.: 1991, Farming, Fertilizers and the NitrateProblem, CAB International, Wallingford, UK.
Allan, J. D.: 1997, Stream Ecology: Structure and Function of Running Waters, Chapman & Hall,London, 388 pp.
APHA-AWWA-WPCF: 1995, Standard Methods for the Examination of Water and Wastewater (18thedn), Greenberg A.H., Clesceri, I.S. and Eaton, A.D (eds), American Public Health, 1008 pp.
Bordalo, A. A., Nilsumranchit, W. and Chalermwat, K.: 2001, ‘Water quality and uses of the Bang-pakonk River (Eastern Thailand)’, Water Res. 35(15), 3635–3642.
Brown, R. M., McClelland, N. I., Deininger, R. A. and Tozer, R. G.: 1970, ‘A water quality index: Dowe dare?’, Water Sewage Works 117, 339–343.
Chapra, S.: 1997, Surface Water Quality Modelling, McGraw-Hill Series in Water Resources andEnvironmental Engineering, McGraw-Hill, New York, 844 pp.
Conesa, V.: 1997, Methodological Guide for Environmental Impact Evaluation (Guıa metodologicapara la evaluacion del impacto ambiental), Ediciones Mundi-Prensa, Madrid, 390 pp.
Cude, C.: 2001, ‘Oregon water quality index: A tool for evaluating water quality management effec-tiveness’, J. Am. Water. Resour. Assoc. 37(1), 125–137.
Emsley, J.: 1994, The Consumer’s Good Chemical Guide, W.H. Freaman Spectrum, Oxford, pp.204–233.
Elser, J. J., Marzolf, E. R. and Goldman, C. R.: 1999, ‘Phosphorus and nitrogen limitation of phy-toplankton growth in the freshwater of North America: A review and critique of experimentalenrichments’, Can. J. Fishes Aquatic Sci. 47, 1468–1477.
Flores, J. C.: 2002, ‘Comments on the use of water quality indices to verify the impact of CordobaCity (Argentina) on Suquia River’, Water Res. 36(18), 4664–4666.
Hallock, D.: 2002, A Water Quality Index for Ecology’s Stream Monitoring Program, Technical Re-port, P No. 02-03-52, Washington Department of Ecology, Environmental Assessment Program,Olympia, WA.
Hatch, D., Goulding, K. and Murphy, D.: 2002, ‘Nitrogen Pollution Causing Problems’, in: P. M.Haygarth and S. C. Jarvis (eds), Agriculture, Hydrology and Water Quality, CABI Publishing,Wallingford, UK, pp 18–19.
Hellawell, J. M.: 1986, Biological Indicators of Freshwater Pollution and Environmental Management,Elsevier, England, 546 pp.
Horton, R. K.: 1965, ‘An index number system for rating water quality’, J. Water Pollut. ControlFeder. 37(3), 300–306.
Jonnalagadda, S. B. and Mhere, G.: 2001, ‘Water quality of the Odzi River in the eastern highlandsof Zimbabwe’, Water Res. 35(10), 2371–2376.
Margalef, R.: 1983, in: S.A. Omega (ed), Limnologıa, Barcelona, Espana, 1010 pp.
322 P. DEBELS ET AL
Martınez de Bascaran, G.: 1979, ‘Establishment of a methodology for the determination of water qual-ity (Establecimiento de una metodologıa para conocer la calidad del agua)’, Boletın InformativoMedio Ambiente 9, 30–51, CIMA, MOPU, Madrid.
Miller, W. W., Joung, H. M., Mahannah, C. N. and Garrett, J. R.: 1986, ‘Identification of water qualitydifferences in Nevada through index application’, J. Environ. Quality 15, 265–272.
Nagels, J. W., Davies-Colley, R. J. and Smith D. G.: 2002, ‘A water quality index for contact recreationin New Zealand’, Water Sci. Technol. 43, 285–292.
Ott, W. R.: 1978, Environmental Indices: Theory and Practice, Ann Arbor Science Publishers Inc.,Ann Arbor, Michigan.
Parra, O., Valdovinos, C., Habit, E. and Figueroa R.: 2004, Programa de Monitoreo de la Calidaddel Agua del Sistema Rıo Biobıo, Technical Report, Center for Environmental Sciences EULA-CHILE, University of Concepcion, Concepcion, Chile.
Pesce, S. F. and Wunderlin, D. A.: 2000, ‘Use of water quality indices to verify the impact of Cordobacity (Argentina) on Suquıa River’, Water Res. 34(11), 2915–2926.
Pesce, S. F. and Wunderlin, D. A.: 2002, ‘Reply to comment on use of water quality indices to verifythe impact of Cordoba City (Argentina) on Suquia River’, Water Res. 36(19), 4990–4991.
Reddy, K. R. and Tucker, J. C.: 1983, ‘Effect of nitrogen source on productivity and nutrient uptakeof water hyacinth (Eicchornia crassipes)’, Econ. Bot. 37, 236–246.
Sastre, A.V., Santinelli, N. H., Otano, S. H. and Ivanissevich, M. E.: 1998, ‘Water quality in thelower section of the Chubut River, Patagonia, Argentina’, Verhandlundgen der internationaleVereinigung fur theoretische und angewandte Limnologie 26, 951–955.
Sharpley, A. N., McDowell, R. W. and Kleinman, P. J.: 2001, ‘Phosphorus loss from land to water:integrating agricultural and environmental management’, Plant Soil 237, 287–307.
Soulsby, C., Langan, S. J. and Neal, C.: 2001, ‘Environmental change, land use and water quality inScotland: current issues and future prospects’, Sci. Total Environ. 265, 387–394.
Stambuck-Giljanovic, N.: 1999, ‘Water quality evaluation by index in Dalmatia’, Water Res. 33(16),3426–3440.
Stanley, D. W. and Hobbie, J. E.: 1981, ‘Nitrogen recycling in a North Carolina coastal river’, Limnol.Oceanogr. 26, 30–42.
Ward, R.: 2001, ‘Development and use of water quality criteria and standards in the United States’.Regional Environ. Change 2(2), 66–72.
Zagatto, P. A., Lorenzetti, M. L., Perez, L. S. N., Menegon, Jr. N. and Buratini, S. V.: 1998, ‘Proposalfor a new water quality index’, Verhandlundgen der internationale Vereinigung fur theoretischeund angewandte Limnologie 26, 2449–2451.