Science of the Total Environment 328 (2004) 195–206

Water quality change in reservoirs of Shenzhen, China: detection

using LANDSATyTM data
Yunpeng Wang*, Hao Xia, Jiamo Fu, Guoying Sheng
State Key Laboratory of Organic Geochemistry, Guangzhou Institute of Geochemistry, Chinese Academy of Sciences,
P.O. Box 1131, Guangzhou 510640, PR China

Received 8 April 2003; received in revised form 17 November 2003; accepted 13 February 2004

Abstract

The objective of this research is to explore a precise and fast way of monitoring water chemical and biochemical
quality in the reservoirs of Shenzhen, China. Water quality change in 1988 and 1996 are detected by synthesizing
satellite data and ground-based data. One scene Thematic Mapper (TM) image in winter of 1996 was acquired and
the simultaneous in situ measurement, sampling and analysis were performed. Main methods include radiometric
calibration of TM remote sensor, atmospheric correction to image data and statistical model construction. The results
indicate that satellite-based estimates and in situ measured water reflectance have very high correlation, and the root
mean square differences between two kinds of indices are close to 0.02–0.03 for each TM band in Visible-Near
Infrared (VI-NIR) range. Statistical relationship between calibrated image data (average of 5=5 pixels) of TM bands
and laboratory analyzed data of water samples indicated reflectance of TM band 1 to band 4 and organic pollution
measurements such as TOC, BOD and COD had higher correlation. The same scene TM data in the winter of 1988
was processed in the same procedure. Results indicate that water quality of most reservoirs have become worse.
Water of eastern reservoirs near Dongjiang River is characterized with higher TOC and TSS, and water of western
reservoirs is characterized with higher BOD and COD.
䊚 2004 Elsevier B.V. All rights reserved.

Keywords: Water quality change; Remote sensing; Landsat-5 TM image; Reservoir water; Shenzhen; Biochemical oxygen demand;
Chemical oxygen demand

1. Introduction Most reservoirs accumulate the runoff formed of

Reservoir water is the main source of drinking
water, industrial water and agricultural water in
Shenzhen and Hong Kong. There is an important
need for ongoing water quality monitoring to
ensure the water standard continuous to be active.
*Corresponding author. Tel.: q86-20-8290170; fax: q86-
20-85290706.
E-mail address: wangyp@gig.ac.cn (Y. Wang).
rainwater, such as Tiegang reservoir and Shiyan
reservoir, and some reservoirs accumulate the
water from rivers with long distance drainage,
such as Shenzhen reservoir whose main water is
from the Dongjiang River. Shenzhen is one area
with the most rapid development speed of econo-
my and urbanization of China in the past 20 years.
Local industrial developments and population
increases have affected reservoir ecosystem and

0048-9697/04/$ - see front matter 䊚 2004 Elsevier B.V. All rights reserved.
doi:10.1016/j.scitotenv.2004.02.020
196 Y. Wang et al. / Science of the Total Environment 328 (2004) 195–206
water quality. Now, and in the future, there isywill
be the need for more high quality water.
Traditional water quality monitoring depends on
in situ measurements or sequent laboratory analysis
of the samples. This kind of point sampling meth-
ods may give accurate measurements, but they are
time and money consuming. Further and most
importantly, they can’t give the real-time spatial
overview that is necessary for the global assess-
ment and monitoring of water quality (Brivio et
al., 2001).
Satellite remote sensing may provide suitable
ways to integrate limnological data collected from
traditional in situ measurements. Since the 1980s,
with improvement of sensor spatial and spectral
resolution, satellite remote sensing has been used
to monitor inland water by using correlation
between broad-band reflectance and other proper-
ties of the water column, including Secchi disk
depth, chlorophyll concentrations, pigment load,
total suspended sediments, temperature and water
quality data analyzed in a laboratory (Schiebe et
al., 1992; Dekker and Peters, 1993; Schneider and
Mauser, 1996; Zilioli and Brivio, 1997; Fraser,
1998; Giardiano et al., 2001; Kloiber et al., 2000,
2002).
The Landsat-5 TM (Thematic Mapper) images
were selected to acquire broadband reflectance
data. All bands of TM Band 1 (TM1) to TM Band
7 (TM7) were used, and the most common were
TM1-TM4 and TM6. TM1-4 bands known as
visible-near infrared bands are in the spectral range
where light passed through the water, providing
some information about water quality (Giardiano
et al., 2001; Kloiber et al., 2000, 2002).
Previous studies prove TM1-4 bands may reflect
the routine parameters of water quality including
aquatic humus, chlorophyll-a, phytoplankton, dis-
solved organic matter, suspended matter, etc. (Dek-
ker et al., 1992; Lathrop, 1992; Dekker and Peters,
1993; Lavery et al., 1993; Pattiaratchi et al., 1994;
Cox et al., 1998; Brivio et al., 2001; Giardiano et
al., 2001; Stadelmann et al., 2001; Kloiber et al.,
2002). In this paper, a study of statistical model
between reflectance of TM band 1–4 and chemical
measurements of water quality including total
organic carbon (TOC), chemical oxygen demand
(COD) and biochemical oxygen demand (BOD)
is presented for the reservoirs of Shenzhen. In
doing so, the paper provides the first information
on the statistical model between water chemical
and biochemical measurements and the satellite
estimated reflectance of TM bands and the prelim-
inary application of the water quality change in
the reservoirs of Shenzhen.
2. Study area
The study area is in the north of Shenzhen,
Guangdong Province, south of China (Fig. 1).
This is a tropical area with abundant rainfall.
According to the Shenzhen meteorological station,
the annual precipitation of Shenzhen varies from
1600–2000 mm, average annual precipitation is
1933.3 mm, and in most area of Shenzhen the
average annual precipitation is over 1700 mm
(http:yywww.szlib.gov.cnyszglyqixiangyqixiang.
htm). There are more than 30 reservoirs in Shen-
zhen and they accumulate water throughout year.
The five largest reservoirs Tiegang, Shiyan, Xili,
Yantian and Shenzhen, which account for over
80% water column of the surface water in Shen-
zhen, are studied here. The typical characteristics
of these reservoirs are the heavy organic pollution
and eutrophication, and the main pollutants include
total nitrogen and phosphorous, BOD and volatile
hydroxybenzene (Annul Report of Shenzhen Envi-
ronmental Protection Bureau, 2002).
3. Material and methods
One scene TM image was acquired and the
simultaneous in situ spectral measurement, water
sampling and laboratory analysis were performed.
Image process methods include radiometric cali-
bration of TM remote sensor and atmospheric
correction to image data. Multiple regression meth-
ods were used for the statistical model construc-
tion. The water quality change was detected by
comparing two scenes of TM data through the
same procedures.
3.1. Satellite data and spectral measurement
The date of the TM image is for the 10 March
1996, a time when the composition of reservoir
water is relatively stable in the dry season. This
date is chosen because the image is cloud-free and
 Y. Wang et al. / Science of the Total Environment 328 (2004) 195–206
Fig. 1. Location of study area, Shenzhen, Guangdong, South of China.
in good quality (Detailed in http:yy
www.rsgs.ac.cn). For comparing the radiation of
water in remote sensor and the in situ measure-
ment, a digital count and related parameters of the
image were used. A sub-image in size of
1200=700 pixels was extracted, which covered
the most of the study area.
For determining the relationship between the
satellite data and in situ measurement, the surface
radiation of water was measured during satellite
overpass. The in situ measurements were taken
using a HG-100 portable spectroradiometer
designed by our institute. It is a spectrally based
colorimeter, which can perform complete radio-
metric measurements in the range of 400–860 nm
with a spectral bandwidth of 10–2 nm accuracy.
This instrument can mount different optics (cor-
responding to 18) and remote cosine receptor
(RCR) fields of view (FOV). A sun-photometer
was used to measure the atmospheric transmittance
in different wavelengths from visible to near infra-
red range.
3.2. Sampling and analysis
The sample gathering sites were arranged to
coincide with satellite passages along a section in
the north–south direction. Sampling number varies
197
from 8 to 10 according to the reservoir size. The
position of the sampling ship was geolocated by a
portable Global Position System (GPS: Model
GM-100 of Beijing InterGPS Corporation). At
each sampling site, water was collected with a
dark-colored bottle from depths of 0 and 2 m.
Then an integrated water sample of two depths
was taken in situ for the further analysis in the
laboratory. The interval between sampling posi-
tions was maintained as short as possible, and
these time was usually between 30 and 60 min. In
the field, all the sample were stored in a tempera-
ture- maintaining (5"2 8C) container with a black
cover to avoid sample deterioration from biological
activity. At the laboratory, all sample bottles were
stored in the refrigerator at the same temperature
prior to analysis.
Laboratory analysis includes TOC, COD and
BOD. All analytical methods used were China
Standard Method (GB Method), which were as
follows:
1. TOC: Determination of TOC by non-dispersive
infrared absorption method (GByT 13 193-
1991).
2. COD: Determination of the chemical oxygen
demand, Dichromate method (GByT 11 914-
1989).
198 Y. Wang et al. / Science of the Total Environment 328 (2004) 195–206

3. BOD: Determination of biochemical oxygen Table 2

demand after 5 days, Dilution and seeding The atmospheric transmittance Tlx values in TM bands meas-
ured by a sun-photometer (March 10, 1996)
method (GByT 7488-1987).
Band 1 2 3 4
The other determination methods were also GB
Methods, which would not be listed here. Tlx 0.6705 0.7433 0.8402 0.9905

3.3. Calibration and atmospheric correction

For comparing the spectral properties of water
received by satellite and the in situ measurements,
the satellite image data needed to be calibrated
and corrected for atmospheric effects. The objec-
tive of absolute radiometric sensor calibration of a
satellite is to produce accurate calibrated data in
SI units for broad usage (Brivio et al., 2001).
Compared to soil and vegetation, the fraction of
light reflected from water is very small (Maul,
1985). Therefore, accurate absolute radiometric
correction of the sensor is critical (Gordon, 1987).
Thome et al. have done the long-term research of
TM and ETM data calibration (Thome et al., 1993;
Thome, 2001 ). In this study, Eq. (1) is used to
calibrate the digital signal levels (DSLs), which
are then converted to (Lambertian) apparent reflec-
tance by Eq. (2) with the coefficients in Table 1.
DSLyDSL0
L*s
G 1994). These techniques are grouped into two
pd2SL*
r*s (2)
TlxE0ms
L* is the radiance (Wmy2sry1mmy1); dS: is
Earth–Sun distance normalized with respect to
mean of 1.0 A.U; ms: is cosine of solar zenith
angle; E0: Exo-atmospheric solar irradiance
(Wmy2mmy1); Tlx: is the atmospheric transmit-
tance along the path from the sun to the ground
surface measured by sun-photometer simultaneous-
ly while the satellite over-passing, which is shown
in Table 2.
The haze effect is thought to be the most
important atmospheric effect due to the scattering
and absorption of the radiation by molecules and
aerosols (Kaufman, 1989). Many atmospheric haze
removal techniques have been developed for the
use of digitally remote-sensed data (Kaufman and
Sendra, 1988; Caselles and Lopez Garcia, 1989;
Richter, 1990; Fraser et al., 1992; Gilabert et al.,
(1)
classes: a simple dark object subtraction (DOS)

Table 1
Radiometric calibration coefficients and exo-atmospheric solar irradiances from Landsat-5 TM (Thome et al. (1993) and Palmer
(1984))

TM band Center wavelength Gain, G Offset, DSL0 Solar Irradiance, E0

(mm) (countsyWmy2sry1mmy1) (counts) (Wmy2sry1mmy1)
1 0.4863 1.6599 2.523 1959.2
2 0.5706 0.85099 2.417 1827.4
3 0.6607 1.2411 1.452 1550.0
4 0.8382 1.2277 1.854 1040.8
5 1.677 9.2526 3.423 220.75
6 2.223 17.550 2.633 74.960
1 Gs(y7.84=10y5)(days since launch)q1.409
2 Gs(y2.75=10y5)(days since launch)q0.7414
3 Gs(y1.96=10y5)(days since launch)q0.9377
4 Gs(y1.10=10y5)(days since launch)q1.080
5 Gs(7.88=10y5)(days since launch)q7.235
7 Gs(7.15=10y5)(days since launch)q15.63
 Y. Wang et al. / Science of the Total Environment 328 (2004) 195–206
Fig. 2. Scatter plot of satellite estimated reflectance vs. in situ
measured reflectance of water in three TM bands (42 sites of
reservoirs, Shenzhen).
method and using atmospheric transmission model
combining with in situ field measurements.
Hence, an improved dark-object subtraction
technique for atmospheric haze correction is used
based on the method of Chavez and Teillet after
comparing different atmospheric correction meth-
ods (Chavez, 1988, 1996; Teillet and Fedosejevs,
1995). Eq. (3) is used to compute apparent reflec-
tances at the sensor form DSLs extracted from the
imagery for the dark target and average back-
ground:
pd2SŽDSL-DSL0.
rag*s
E0msGTlx
This equation is used to obtain rag* values for
the dark target DSLs and the average background
DSLs in each of Landsat TM band 1, 2, 3 and 4.
The radiometric calibration gain and offset coeffi-
cients, G and DSL0, listed in Table 1 were deter-
mined by Thome et al. (1993) and Palmer (1984).
E0 is the exo-atmospheric solar irradiance. The
recommended gain coefficients are given in the
lower part of Table 1. A sun angle computation
will give ms for the scene center and ds for the
date of image acquisition by the remote sensor.
199
4. Results
4.1. Satellite estimated and in situ measured reflec-
tance of water
A 5=5 pixel window, corresponding to the area
of 150=150 m, was extracted from the images for
each of the locations in each reservoir. Average
DLS of 25 pixels within the window was converted
to radiance L*, reflectance rag*. Also, the ground
reflectance (rg*) was calculated by DOS methods
as the above description.
Satellite estimated and in situ measured reflec-
tance of water in 42 sites are compared in a scatter
plots in Fig. 2. The results show that the computed
values are close to the measured values, in general,
and present a good fit (Fig. 2). The correlation
analysis results show that the correlation coeffi-
2
cient square (RTM1 ) for TM1 of 42 samples is
0.638, 0.607 for TM2 and 0.780 for TM3. The
satellite-based estimates and in situ measured water
reflectance are significantly correlated: the t-sig-
nificance test of correlation analysis was used (ts
8.18)t0.01(40)s2.704, P-0.01).
For assessing the absolute error of this method,
the root mean square (RMS) errors in TM band
1–4 for the 42 sites were computed. RMS is the
standard deviation of the differences between sat-
ellite estimated and in situ measurements of water
reflectances. The bar graph of TM band 1–4 is
presented in Fig. 3. It shows that on average, the
root mean square differences between the two
kinds of indices are close to 0.02–0.03 in VI-NIR
(3)
Fig. 3. The root mean square errors of water reflectance dif-
ferences in TM band 1–4 between satellite and ground meas-
urement, 42 locations in Shenzhen.
200 Y. Wang et al. / Science of the Total Environment 328 (2004) 195–206

TOCs6.41y85.29r1y2.05r2y24.96r3

(MRs0.829, Ns40,

standard error of estimates0.25) (4)

BODs1.79y0.789r1q52.36r2y3.28r3

(MRs0.707, Ns40,

standard error of estimates0.24) (5)

CODs2.76y17.27r1q72.15r2y12.11r3

(MRs0.626, Ns40,

standard error of estimates0.30) (6)

Fig. 4. Scatter plot of predicted values and analyzed values of
TOC (40 samples, rs0.911, Ps0.01).
range. It indicates that satellite estimated reflec-
tance of water presents a high accuracy in TM
band 1–4.
4.2. Water quality models
Where r1, r2 and r3 are the atmospherically
corrected reflectance of TM band 1–3. MR is the
multiple linear correlation coefficient and N is
sample number.
Scatter plots of predicted values from the Eqs.
(4)–(6) vs. analyzed values of TOC, BOD and
COD are shown in Figs. 4–6. The correlation
coefficients are 0.911(Ps0.01) for TOC,
0.841(Ps0.01) for BOD and 0.791(Ps0.01) for

Statistical techniques were used to determine

the relationships between the satellite estimated
reflectances or their ratios and water quality para-
meters just like in previous studies (Baban, 1993;
Mayo et al., 1995; etc.). A few previous studies
used non-linear power models to address the cur-
vilinear behavior of this relationship (Lathrop,
1992; Cox et al., 1998). In this study, some
multiple regression methods including linear, expo-
nential and log transformations were used to exam-
ine the statistical models between water parameters
analyzed in laboratory (TOC, BOD, and COD)
and remote sensing reflectance values in TM 1–4.
Through comparing, we chose the multiple linear
regression method because of its higher multiple
correlation coefficient. The regression equations of
TOC, BOD and COD and the atmospherically
corrected reflectance of TM band 1–3 (r1, r2 and Fig. 5. Scatter plot of predicted values and analyzed values of
r3) were as follows: BOD (40 samples, rs0.841, Ps0.01).
 Y. Wang et al. / Science of the Total Environment 328 (2004) 195–206
Fig. 6. Scatter plot of predicted values and analyzed values of
COD (40 samples, rs0.791, Ps0.01).
COD. The results indicate that reflectance of TM
band 1–3 and organic pollution measurements
such as TOC, BOD and COD have relatively
higher correlation. Eqs. (4)–(6) are proved effec-
tive to be used in predicting the water quality
parameters of TOC, BOD and COD.
4.3. Water quality change detection in 1988 and
1996
The same scene TM data in the same season of
1988 (January 10) was processed in the same
procedure on the basis of radiometric calibration
and atmospheric correction. The calibration and
atmospheric correction is of critical importance for
the multi-temporal remote sensing data while using
in change detection. For verifying the result of the
calibration and atmospheric correction methods, A
Scatter plot of atmospherically corrected reflec-
tance of TM1 vs. TM2 for some dark objects such
as water, shadow and dense vegetation in 1988
and 1996 was constructed, showing in Fig. 7. The
reflectance (i.e. TM1 vs. TM2) of these dark
objects in different years (1988–1996 with the
interval of 8 years) is distributed in the same areas
generally (Fig. 7). The distributions of shadow
and dense vegetation are very close in 1988 and
201
1996, while the distribution of water is relatively
dispersive. It is due to the change of water quality.
The results show that calibration and atmospheric
correction to TM data is effective.
Regression Eqs. (4)–(6) were applied to the
whole surface of the reservoirs and produced the
density-sliced map of TOC, BOD and COD. These
measurements are used for water classification
according to the Standard of State Environment
Protection Bureau, China. Fig. 8 is a comparison
of two BOD maps of Shiyan reservoir in 1988 and
1996, which were calculated by Eq. (5).
The BOD value varies from 2 to 3 mg ly1 in
1988, and the high BOD area locates in the middle
to northern parts of the reservoir (Fig. 8). How-
ever, in 1996, BOD value varies from 2.5 mg ly1
to more than 3.5 mg ly1, and the BOD value in
most of the reservoirs is larger than 3 mg ly1. The
map of COD shows the same characteristic in
space distribution. It indicates the pollution of
Shiyan reservoir has increased from 1988 to 1996.
The other reservoirs in the western area of Shen-
zhen show the same trend, and the main source of
the pollution is from the agriculture pollution.
Comparison of two TOC maps of five reservoirs
in 1988 and 1996 (Fig. 9) reveals that TOC value
varies from 2 to 3.5 mg ly1 in 1988, and the high
TOC areas locate in the Northern parts of Shiyan
reservoir and the southern part of Shenzhen reser-
voir. However, in 1996, TOC value in all reservoirs
Fig. 7. Scatter plot of atmospherically corrected reflectance of
TM1 vs. TM2 for water, shadow and vegetation (1988 and
1996)
202 Y. Wang et al. / Science of the Total Environment 328 (2004) 195–206

Fig. 8. BOD maps of Shiyan reservoir in 1988 (up) and 1996 (bottom), calculated from TM atmospherically corrected reflectance
through statistical model (Eq. (5)).

were higher, varying from 3.5 mg ly1 to more
than 5 mg ly1, and the TOC value in most of the
reservoirs was larger than 4 mg ly1. The charac-
teristic in space distribution indicates that the
pollution of all reservoirs in Shenzhen have
increased from 1988 to 1996. With the reservoirs
in the eastern area of Shenzhen showing relatively
worse trend, as the source of the pollution is from
increased agriculture, industry and urban pollution.
5. Conclusions
A precise and fast way of water quality moni-
toring was explored in some reservoirs of Shen-
 Y. Wang et al. / Science of the Total Environment 328 (2004) 195–206 203

Fig. 9. TOC maps of reservoirs in Shenzhen in1988 (up) and 1996(bottom), calculated from TM atmospherically corrected reflec-
tance through statistical model (Eq. (4)). 1. Shiyan Reservoir; 2. Tiegang Reservoir; 3. Xili Reservoir; 4. Yantian Reservori; 5.
Shenzhen Reservoir.
204 Y. Wang et al. / Science of the Total Environment 328 (2004) 195–206
zhen. Water quality change in 1988 and 1996 are
detected by remote sensing. One scene TM image
in winter of 1996 was acquired with simultaneous
in situ measurement, sampling and analysis. The
radiometric calibration of TM remote sensor and
atmospheric correction to image data and statistical
model construction were performed. The results
indicate that satellite-based estimates and in situ
measured water reflectance have very high corre-
lation, and the root mean square difference
between two kinds of indices are close to 0.02–
0.03 for each TM band in VI-NIR range.
The study shows that satellite estimated reflec-
tance of water presents a high accuracy in TM
band 1–4. Correlation between calibrated image
data (average of 5=5 pixels) of TM bands and
laboratory analyzed data of water samples were
performed and results indicate that reflectance of
TM band 1 to band 4 and TOC, BOD and COD
have higher correlation. The statistical relationship
between satellite-based reflectance and organic
pollution measurements including TOC, BOD and
COD were constructed. The standard errors of
estimation of these models vary from 0.24 to 0.30.
The same scene TM data in the same season of
1988 was processed in the same procedure on the
base of calibration and atmospheric correction to
the remote sensed data. Results indicate water
quality of most reservoirs has become worse.
Water of eastern reservoirs near Dongjiang River
is characterized with higher TOC and TSS, and
water of western reservoirs is characterized with
higher BOD and COD.
This study demonstrates the possibility, accura-
cy, potential and effectiveness for determining the
water quality measurements of reservoir such as
TOC, BOD and COD with a little ground proof.
Hence this technique has many advantages, such
as more convenient, providing the spatial distri-
bution of water parameters and relatively less time
and money consuming comparing with traditional
methods. Some factors must be considered while
this technique will be used. These factors include:
● A composite sample from relative large areas
in one location should be used for the labora-
tory analysis instead of a single sample, because
the spatial resolution of TM.
● The size of the location pixels should be eval-
uated. Of course, single pixel DSL analyzed
could not be compared with a ground sample
because of the fluidity of the water. Some
experts take the scale of 3=3 pixels (Giardiano
et al., 2001). According to our experience,
larger scales should be considered, such as
5=5 pixels in this research.
● In situ measurements of radiation and atmos-
pheric conditions are vitally important both for
the calibration and correction to remotely
sensed data and the statistical model construc-
tion of water quality. In a new area, it is
necessary to perform the in situ measurements
of spectra and atmospheric conditions, even
through the same technique and procedure are
used. Combining with the actual measurement,
different methods of calibration and atmospher-
ic correction should be compared in a new area
or reservoir to determine the better strategy.
● The complexity of water quality affects the
accuracy of this technique. TSS should be taken
account in, especially in the shallow area of
the water. Seasonal variation is another impor-
tant factor to the remote sensor and hydraulic
conditions. The calibration parameters should
be adjusted in different seasons. This study is
undertaken in dry season, the calibration para-
meters and statistical models should be modi-
fied while being used in humid season.
● Multi-regression method should be proved and
modified by some ground sample analysis
results and applied regionally. Finally, although
the satellite data can be used to reflect the
water quality parameters such as Secchi disk
depth, chlorophyll concentrations, pigment
load, total suspended sediments, temperature
and some chemical properties, and this tech-
nique is valuable and important for remote
areas where direct access is not easy and where
the sum of sampling and analytical chemistry
cost is high, it should be emphasized that this
technique cannot substitute the traditional meth-
ods because some parameters of water quality
such as heavy metals, nitrate, phosphate and
 Y. Wang et al. / Science of the Total Environment 328 (2004) 195–206
organic pollutants cannot be determined by
remote sensing.
Acknowledgments
This work was supported by National Natural
Science Foundation of China (Grant No.
40171077, 49901014), National Key Basic
Research Programming Project (Grant No.
2001CB209101) and Natural Science Foundation
of Guangdong Province (Grant No. 010505). We
want to greatly thank Kurtis J. Thome and Magaly
Koch for providing some materials of TM data
calibration and Kim Neth for improving English
of this paper. We also want to thank two anony-
mous reviewers for their constructive comments.
References
Baban SMJ. Detecting water quality parameters in the Norfolk
Broads, UK, using Landsat imager. Int J Remote Sens
1993;14:1247 –1267.
Brivio PA, Giardino C, Zilioli E. Validation of satellite data
for quality assurance in lake monitoring applications. Sci
Total Environ 2001;268:3 –13.
Caselles V, Lopez Garcia MJ. An alternative simple approach
to estimate atmospheric correction in multitemporal studies.
Int J Remote Sens 1989;10:1127 –1134.
Chavez PS. An improved dark-object subtraction technique for
atmospheric scattering correction of multispectral data.
Remote Sens Environ 1988;24:459 –479.
Chavez PS. Image-based atmospheric corrections – revised
and improved. PhotoGramm Eng Remote Sens
1996;62:1025 –1036.
Cox RM, Forsythe RD, Vaughan GE, Olmsted LL. Assessing
water quality in the Catawba River reservoirs using Landsat
Thematic Mapper satellite data. Lake Reservoir Manage
1998;14:405 –416.
Dekker AG, Malthus TJ, Wijnen MM, Seyhan E. The effect
of spectral bandwidth and positioning on the spectral sig-
nature analysis of inland waters. Remote Sens Environ
1992;41:211 –225.
Dekker AG, Peters SWM. The use of the thematic mapper for
the analysis of eutrophic lakes: a case study in the Nether-
lands. Int J Remote Sens 1993;14:788 –821.
Fraser RS, Ferrare RA, Kaufman YJ, Markham BL. Algorithm
for atmospheric corrections of aircraft and satellite images.
Int J Remote Sens 1992;13:541 –557.
Fraser RS. Multispectral remote sensing of turbidity among
Nebraska Sand Hills lakes. Int J Remote Sens
1998;19:3011 –3016.
205
Giardiano C, Pepe M, Brivio PA, Ghezzi P, Zilioli E. Detecting
chlorophyll, secchi disk depth and surface temperature in a
sub-alpine lake using Landsat imagery. Sci Total Environ
2001;268:19 –29.
Gilabert MA, Conese C, Maselli F. An atmospheric correction
method for the automatic retrieval of surface reflectances
from TM images. Int. J Remote Sens 1994;15:2065 –2086.
Gordon HR. Calibration requirements and methodology for
remote sensors viewing the oceans in the visible. Remote
Sens Environ 1987;22:103 –126.
Kaufman YJ, Sendra C. Algorithm for automatic corrections
to visible and near-infrared satellite imagery. Int. J Remote
Sens 1988;9:1357 –1381.
Kaufman YJ. The atmospheric effect on remote sensing and
its correction. Theory and applications of optical remote
sensing. New York: John Wiley and Sons, 1989. p. 336 –
428.
Kloiber SM, Anderle TH, Brezonik PL, Olmanson LG, Bauer
ME, Brown DA. Trophic state assessment of lakes in the
Twin Cities (Minnesota, USA) region by satellite imagery.
Arch Hydrobiol Adv Limnol 2000;55:137 –151.
Kloiber SM, Brezonik PL, Olmanson LG. A procedure for
regional lake water clarity assessment using Landsat multi-
spectral data. Remote Sens Environ 2002;82:38 –47.
Lathrop RG. Landsat thematic mapper monitoring of turbid
inland water quality. Photogramm Eng Remote Sens
1992;58:465 –470.
Lavery P, Pattiaratchi C, Wyllie A, Hick P. Water quality
monitoring in estuarine waters using the Landsat Thematic
Mapper. Remote Sens Environ 1993;3:268 –280.
Maul GA. Introduction to satellite oceanography. Dordrecht:
Maritinus Nijhoff Publisher, 1985.
Mayo M, Gitelson A, Yacobi YZ, Ben-Abraham Z. Chloro-
phyll distribution in lake Kinneret determined from Landsat
Thematic Mapper data. Int J Remote Sens 1995;16:175 –
182.
Palmer JM. Effective bandwidths for Landsat-4 and Landsat
D1 multispectral scanner and thematic mapper subsystem.
IEEE Trans Geosci Remote Sens 1984;GE-22:336 –338.
Pattiaratchi C, Lavery P, Wyllie A, Hick P. Estimates of water
quality in coastal waters using multi-date Landsat Thematic
Mapper data. Int J Remote Sens 1994;15:1571 –1584.
Richter RA. Fast atmospheric correction algorithm applied to
Landsat TM images. Int J Remote Sens 1990;11:159 –166.
Schiebe FR, Harrington JA, Ritchie JC. Remote sensing of
suspended sediments: the Lake Chicot, Arkansas project. Int
J Remote Sens 1992;13:1487 –1509.
Schneider K, Mauser W. Processing and accuracy of Landsat
thematic mapper for lake surface temperature measurement.
Int J Remote Sens 1996;11:2027 –2041.
Stadelmann TH, Brezonik PL, Kloiber SM. Seasonal patterns
of chlorophyll-a and Secchi disk transparency in lakes of
east–central Minnesota: implications for design of ground-
and satellite-based monitoring programs. Lake Reservoir
Manage 2001;17:299 –314.
206 Y. Wang et al. / Science of the Total Environment 328 (2004) 195–206

Teillet PM, Fedosejevs G. On the dark target approach to
atmospheric correction of remotely sensed data. Can J
Remote Sens 1995;21:374 –388.
Thome KJ, Gellman DI, Parada RJ, Biggar SF, Slater PN,
Moran MS. In-flight radiometric calibration of Landsat-5
Thematic Mapper form 1984 to present. SPIE
1938;1993:126 –130.
Thome KJ. Absolute radiometric calibration of Landsat 7
ETMq using the reflectance-based method. Remote Sens
Environ 2001;78:27 –38.
Zilioli E, Brivio PA. The satellite derived optical information
for the comparative assessment of lacustrine water quality.
Sci Total Environ 1997;196:229 –245.