www.elsevier.com/locate/marpolbul

Marine Pollution Bulletin 54 (2007) 1267–1272

Remotely sensed assessment of water quality levelsin the Pearl River Estuary, China

Chuqun Chen a,*, Shiling Tang a, Zhilin Pan a, Haigang Zhan a,Magnus Larson b, Lennart Jonsson b

a Key Laboratory of Tropical Marine Environmental Dynamics, South China Sea Institute of Oceanology, Chinese Academy of Sciences,

Guangzhou 510300, Chinab Department of Water Resources Engineering, Lund University, Lund, Sweden

Abstract

In this paper, a method of assessing water quality from satellite data is introduced. The composite pollution index (CPI) was calcu-lated from measured chemical oxygen demand (COD) and nutrient concentration. The relationships between CPI and 240 band com-binations of SeaWiFS water-leaving radiance were analyzed and the optimal band combination for estimating CPI was chosen fromthe 240 band combinations. An algorithm for retrieval of CPI was developed using the optimal band combination, (L443 · L510)/(L412 + L490). The CPI was estimated from atmospherically corrected SeaWiFS data by employing the algorithm. Furthermore, theCPI value range for each water quality level was determined based on data obtained from 850 samples taken in the Pearl River Estuary.The remotely sensed CPIs were then transferred to water quality levels and appropriate maps were derived. The remotely sensed waterquality level maps displayed a similar distribution of levels based on in situ investigation issued by the State Ocean Administration,China. This study demonstrates that remote sensing can play an important role in water quality assessment.� 2007 Elsevier Ltd. All rights reserved.

Keywords: Remote sensing; Assessment; Water quality level; Pearl River Estuary

1. Introduction

Because of population growth and rapid industrializa-tion and urbanization, more and more solid, liquid andgas wastes have been emitted into the environment ofChina. In the Pearl River Delta especially, environmentalpressures have increased substantially during the last twodecades. Environmental conditions have become a majorfactor limiting progress in terms of economic development.As solid and gas wastes will (at least partly) convert toliquid wastes via eluviation or rain, they may influencewater quality. Water quality, as related to solid, liquid,and gas wastes, represents a comprehensive index of envi-

0025-326X/$ - see front matter � 2007 Elsevier Ltd. All rights reserved.

doi:10.1016/j.marpolbul.2007.03.010

* Corresponding author.E-mail address: cqchen@scsio.ac.cn (C. Chen).

ronmental quality. When implementing an environmentprotection policy, it is important to monitor the aquaticenvironment and to provide water quality informationfor government and environmental protection agencies.

Conventional methods of water quality monitoringdepend on in situ measurements and sequential laboratoryanalysis of the samples. These point-sampling methodsmay give accurate measurements of water quality parame-ters, but they are expensive, time-consuming and, mostimportantly, they cannot provide synchronous water qual-ity data for large areas. The remote sensing technique isexpected to be an alternative method of water quality mon-itoring due to its advantages regarding data synchronicity,low cost, and large spatial coverage.

With the development of remote sensing techniques,many studies on the application of satellite data for waterquality monitoring have been reported upon (e.g., Lavery

1268 C. Chen et al. / Marine Pollution Bulletin 54 (2007) 1267–1272

et al., 1993; Pattiaratchi et al., 1994; Oron and Gitelson,1996; Brivio et al., 2001; Ostlund et al., 2001; Hu et al.,2004; Wang et al., 2004). However, most remotely sensedmonitoring investigations have focused on determining sin-gle water quality variables, such as suspended sediments,turbidity, transparency or Secchi disc depth, chlorophyll-a, dissolved organic matters, COD and BOD (Tassan,1994; Doerffer and Fischer, 1994; Arenz et al., 1996; Alleeand Johnson, 1999; Giardiano et al., 2001; Chen et al.,2003a; Wang et al., 2004). It is important to determine sin-gle water quality variables. However, it is often moreimportant to obtain collective information on water qual-ity, as it can provide concise information on environmentalconditions overall. With the introduction of the GreenGDP concept, collective information on water quality notonly provides environmental information for the adminis-trative sectors and the public, but can also play a role inGreen GDP calculations. In order to obtain collectiveinformation on water quality, it is crucial to develop anintegrated method for water quality assessment.

In this paper, a method for integrated assessment ofwater quality conditions is described. The composite pollu-tion index (CPI) calculated from measured chemical oxy-gen demand (COD) and nutrient concentration was usedto quantify the pollution situation in the Pearl River Estu-ary. The water quality of the measured samples was dividedinto five levels, according to China’s national water qualitystandards. The relationship between CPI and the five waterquality levels was then analyzed and the thresholds totransfer CPIs to water quality levels were established basedon the relationship derived.

2. Study area

The Pearl River is 2214 km long and has a catchmentarea of 453,690 km2 (Zhao, 1990). It is the largest river insouthern China, and has a mean annual runoff of 326 bil-lion m3, the second largest river in China after the Yangtze.The annual variation in discharge is significant anddepends on the amount of rainfall the catchment receives.From October to April the weather is dominated by thenortheast monsoon and mean rainfall is relatively lowbetween 30 and 40 mm/month (Zhao, 1990; Justesenet al., 1996). During the period from May to Septemberthe monsoon comes from the southwest and warm, humidair is brought up from lower latitudes causing considerablylarger amounts of rainfall (300–400 mm/month). The dis-charge into the Pearl River Estuary during the dry seasonis about 1500 m3/s but this can increase to 20,000 m3/s dur-ing the wet season. The annual discharge of suspended sed-iments is about 87 million tonnes, of which >90% isdischarged in the wet season. The warm-humid climateresults in strong chemical weathering and about 30 milliontonnes of dissolved matter are discharged per year (Zhao,1990). The water is discharged to the South China Seathrough a delta with eight river branches, resulting in acomplicated spatial distribution of the water’s components.

There are several major cities in the delta, includingGuangzhou, Foshan, Shenzhen, and Zhuhai. Over the lasttwo decades, the whole delta area has experienced rapidindustrialization and urbanization. This economic expan-sion has adversely impacted water quality in the down-stream areas of the Pearl River system because ofincreases in the discharge of industrial and domestic waste-waters (Jayawardena and Lai, 1989). Water quality in theestuary has become progressively worse in the last 20 years.

The most serious environmental issue in the Pearl RiverEstuary is eutrophication. Since the nutrients involvedmostly originate from discharged runoff, their concentra-tions are higher in the surface layer than at the bottom, dis-playing a seasonal variation (Yin, 2003). The concentrationsof total inorganic nitrogen can be high (up to 126 lM, andgenerally >70 lM), however, PO4 concentrations are gener-ally low (maximum 1.5 lM, and mostly <1 lM). N:P ratios(N = total inorganic nitrogen, P = PO4) are typically >50:1,although they can rise to >100:1, which is much higher than16:1, the standard nutrient ratio for phytoplankton inges-tion. The nutrient ratio shows that P is the main limiting fac-tor in controlling phytoplankton biomass production.Eutrophication has produced serious red tide events orharmful algal blooms in the estuary and its adjacent coastalwaters since the 1980s causing economic losses. Environ-mental conditions in the area are hence a serious concern.Recovery programmes for water quality are in operationand the trend in water quality degradation in the down-stream waters of the Pearl River and its estuary will hope-fully improve. The remote sensing technique is expected tomonitor such a recovery programmes.

3. Data employed and methodology

3.1. In situ data collection

In situ data were collected during six cruises conductedin the Pearl River Estuary between 12–13 March, 19–20March, 9–10 July, 15–16 December, and 29–31 Decemberin 1998, and from 25 to 26 January 2003. In situ measure-ments were taken at three depths from 14 stations duringthe five cruises in 1998 (the sampling positions are shownin Fig. 1a) and at surface for 18 stations during January2003 (the sampling positions are shown in Fig. 1b).

During the cruises in 1998, a total of 832 water sampleswere analyzed, and the variables measured included tem-perature, depth, salinity, pH, chlorophyll-a, suspended sol-ids (SS), nutrients (N, P, and Si), total organic carbon(TOC), dissolved organic carbon (DOC), chemical oxygendemand (COD), biochemical oxygen demand (BOD5) andthe heavy metals Cu, Pb, Zn, Cr, Cd, Ni and Hg. Duringthe cruise in 2003 a total of 18 water samples were ana-lyzed, the variables measured included temperature, depth,salinity, chlorophyll-a, total suspended matters (TSM),nutrients (N, P, and Si), dissolved organic carbon (DOC),chemical oxygen demand (COD), gelbstoff absorption coef-ficient (Ag), and remote sensing reflectance (Rrs). All the

Fig. 1. The sampling positions of cruises in 1998 (a) and in January 2003 (b) in Pearl River Estuary.

C. Chen et al. / Marine Pollution Bulletin 54 (2007) 1267–1272 1269

analytical methods used were in accord with China’s Stan-dard Method, the Specifications for Oceanographic Sur-veys, Observations of Chemical Parameters in Sea Water(GB 12763.4-91,) and Specifications for Oceanographicand Marine Biological Surveys (GB 12763.6-91).

3.2. Composite pollution index (CPI)

The environmental situation in the Pearl River Estuaryis mainly related to organic pollution and eutrophication.Nitrogen and phosphate are the main nutrients for phyto-

1270 C. Chen et al. / Marine Pollution Bulletin 54 (2007) 1267–1272

plankton growth, and eutrophication depends on theirabundance. Thus, chemical oxygen demand (COD) wasused as an index for organic pollution. The composite pol-lution index (CPI) was utilized to scale the collective pollu-tion state, especially with regard to eutrophication. CPIwas calculated according to the following equation:

CPI ¼ COD� TIN� P� 106

4500;

where COD stands for chemical oxygen demand (mg/L),TIN is the total inorganic nitrogen concentration (mg/L),and P is the phosphate (PO4) concentration (mg/L). ACPI was calculated for the in situ measured data of allwater samples.

3.3. SeaWiFS data processing

The sea-viewing wide field-of-view sensor (SeaWiFS)data used here were received by the SeaWiFS receiving sta-tion, KUS, at the Hong Kong University of Science &Technology, and distributed by NASA Goddard DAAC(Distributed Active Archive Center). The SeaWiFS dataunderwent atmospheric corrections with a local, improvedmethod (Chen et al., 2003b). Two SeaWiFS image scenes,acquired on 31 December 1998, and 29 January 2003, wereprocessed and the water-leaving radiance, Lw, of SeaWiFSband 1–6 were obtained from the SeaWiFS L1A data. Thewater-leaving radiance data were read out from the twoprocessed SeaWiFS images corresponding to the samplinglocations of 30–31 December 1998 and 25–26 January,2003 cruises, respectively.

Table 1The relationship between CPIs and water quality levels

Levels WQL 1 WQL 2 WQL 3 WQL 4 WQL 5

CPI average 0.10907 0.33736 0.58891 0.98598 4.50345N 147 53 34 44 572CPI thresholds <0.22322 >0.22322 >0.46313 >0.78745 >2.74472

<0.46313 <0.78745 <2.74472

N = number of samples.

3.4. Development of an algorithm for CPI estimation

The CPI data measured in 30–31 December 1998 and25–26 January 2003, and the SeaWiFS water-leaving radi-ance data of 31 December 1998, and 29 January 2003 wereused for the development of an algorithm to estimate CPI.This required an establishment of either an optimum bandor band compositions. SeaWiFS Bands 1–6 can generate alarge number of band compositions by operations, such asratio band compositions. Arithmetic and logarithmic oper-ations were used to generate band compositions, and 240 ofthese were considered as candidate compositions. The rela-tionships between all these band compositions and CPIwere analyzed using the in situ CPI data and SeaWiFSwater-leaving radiance data. The analysis showed that theband composition (L2 · L4)/(L1 + L3) had the most signif-icant relationship with the CPI, and this ratio was used fordeveloping an algorithm to estimate CPI. Based on regres-sion analysis, the following expression was obtained:

log ðCPIÞ ¼ 2:8792� ðL2 � L4Þ=ðL1 þ L3Þ � 3:9101

ðN ¼ 30;R ¼ 0:93478Þ;

where L stands for the SeaWiFS water-leaving radianceand subscripts 1–4 denote the SeaWiFS band numbers.

4. Results

4.1. Water quality level (WQL)

CPI values were estimated by application of the CPIalgorithm to the atmospheric corrected SeaWiFS data,such that a CPI distribution image in the Pearl River Estu-ary could be derived. However, as the CPI is not a familiarparameter to the public or to environmental administra-tors, it is necessary to use a more widely used parameterfor water quality assessment. In order to make a remotelysensed assessment of water quality that is generallyaccepted and practical for characterizing the environmentalconditions, the commonly used WQL was employed as aparameter for water quality assessment. China’s NationalStandards on marine water quality (GB3907-1997) havebeen widely used to describe water quality. These waterquality levels are accepted by the public and environmentaladministrators. Water quality may be divided into four lev-els according to the GB3907-1997 standard. In practice,however, if water quality is worse than Level 4, It is termed‘‘worse than level 4’’, so that level can actually be dividedinto five.

In order to transfer the CPI values to water quality lev-els, water samples were classified and related to the four(five) water quality levels of China’s National Standard(GB3907-1997). Of 850 samples from the Pearl River Estu-ary, 147 belonged to water quality Level 1, the clearestwater, and were taken from the southern area of the estu-ary (the offshore areas), However, 572 samples belonged towater quality Level 5, the most polluted water, which weretaken from the estuary close to the river mouth. The rela-tionships between water quality levels and CPIs were ana-lyzed. The average CPI value corresponding to each waterquality level was calculated and is shown in Table 1. Themean value of two adjacent water quality levels was usedas the threshold for separating them.

4.2. Integrated assessment of water quality levels

The CPI was estimated by application of the CPI algo-rithm to the atmospheric corrected SeaWiFS data, andwas used as a temporary variable. CPI values were thentransferred to WQLs values according to the thresholdsshown in Table 1. Water quality level information canbe obtained from satellite remote sensing data via the

Fig. 2. (a) Water quality levels retrieved from SeaWiFS data; (b) waterquality issued by State Ocean Administration, China.

C. Chen et al. / Marine Pollution Bulletin 54 (2007) 1267–1272 1271

CPI-WQL transformation. Fig. 2a is the distribution mapof water quality levels in the Pearl River Estuary and itsadjacent coastal waters as retrieved from SeaWiFS data,and Fig. 2b is a water quality map based on in situ data,issued by the State Ocean Administration (SOA), China.The water quality data in the two figures have a similardistribution pattern. It should be noted that the SOAmap represents an average situation, whereas the satellite-based map refers to a specific time when the satelliteoverpassed.

5. Discussion and conclusion

Water quality levels depend on more than 30 variables,and the main ones of these are different for different watersand can change from place to place. Thus, it is difficult todevelop a common algorithm which is valid for assessmentof all kinds of waters. The remote sensing method forassessment of water quality levels will be changed for differ-ent cases. If the key variable affecting water quality levels istemperature, for instance, in some coastal waters with ther-mal pollution, the remotely sensed method for assessmentof water quality levels is to detect the water temperaturefrom satellite data and then according to the difference ofthe water temperature to the background temperature fordetermination of water quality levels. In the Pearl RiverEstuary, eutrophication is the main environmental prob-lem, so that nutrients and COD were used to calculatethe composite pollution index (CPI) to describe the waterquality conditions, and to identify water quality levels.The water quality levels transferred from the CPI, whichwere retrieved from ocean colour remote sensing data usingthe local algorithm, can be in good accordance with in situinvestigation.

The water-leaving radiance detected by satellite sensorswas the collective radiance contributed by all componentsin the water. The spectral properties of pollutants andnutrients and their influences on water-leaving radianceare not well known. As the pollutants and nutrients inthe Pearl River Estuary mostly originate from the river dis-charge, they should have been subject to similar processes(e.g. mixture and dilution) as for the optically active com-ponents of terrestrial origin (e.g. suspended solids, dis-solved organic matter). This implies that pollutants andnutrients will co-exist with optical active components andcould display a relationship with water-leaving radiance.Based on this relationship, it is possible to assess waterquality from satellite-acquired optical data (e.g. ocean col-our satellite data).

The ocean colour remote sensing data can be theoretic-aly obtain every 1–2 days. However, the ocean colourremote sensing data without cloud-contamination in thisarea are very limited. The satellite data should be processedto monthly or seasonly mean value images to avoid theimpacts from the clouds and to get an integrated satelliteimage for assessment of water quality.

As water quality is affected not only by dissolved pollu-tants but also by solid and gas pollutants via eluviation orrain, water quality could become a collective index for theregional environment. The integrated assessment of waterquality could be useful not only for environmental manage-ment but also for Green GDP calculation.

Acknowledgments

The research work was jointly supported by the ChineseNational 863 Project (2002AA639130), the key project ofGuangdong Natural Science Foundation (06105018), the

1272 C. Chen et al. / Marine Pollution Bulletin 54 (2007) 1267–1272

Innovative Project of Chinese Academy of Sciences(KSCZ2-SW-132), the Swedish Asian-Link CooperationProject, and China–UK scientific foundation. Many thanksto the SeaWiFS Project within NASA GSFC, for providingthe SeaWiFS data and SeaDAS software.

References

Allee, R.J., Johnson, J.E., 1999. Use of satellite imagery to estimatesurface chlorophyll-a and Secchi disc depth of Bull Shoals, Arkansas,USA. International Journal of Remote Sensing 20, 1057–1072.

Arenz Jr., R.F., Lewis Jr., W.M., Saunders, J.F., 1996. Determination ofchlorophyll and dissolved organic carbon from reflectance data forColorado reservoirs. International Journal of Remote Sensing 17 (8),1547–1566.

Brivio, P.A., Giardino, C., Zilioli, E., 2001. Validation of satellite data forquality assurance in lake monitoring applications. The Science of theTotal Environment 268, 3–18.

Chen, C.Q., Shi, P., Zhan, H.G., 2003a. A local algorithm for estimationof yellow substance (gellbstoff) in coastal waters from Sea WiFS data:Pearl River estuary, China. International Journal of Remote Sensing24 (5), 1171–1176.

Chen, C.Q., Wei, J., Shi, P., 2003b. Atmospheric correction of SeaWiFSimagery for turbid waters in Southern China coastal areas. Proceed-ings of SPIE (Ocean Remote Sensing and Applications) 4892, 80–86.

Doerffer, R., Fischer, J., 1994. Concentrations of chlorophyll, suspendedmatter, and gelbstoff in case II waters derived from satellite coastalzone color scanner data with inverse modeling methods. Journal ofGeophysics Research C4 99, 7457–7466.

Giardiano, C., Pepe, M., Brivio, P.A., Ghezzi, P., Zilioli, E., 2001.Detecting chlorophyll, secchi disk depth and surface temperature in asub-alpine lake using Landsat imagery. The Science of the TotalEnvironment 268, 19–29.

Hu, C.M., Chen, Z.Q., Clayton, T.D., Swarzenski, P., Brock, J.C., Muller-Karger, F.E., 2004. Assessment of estuarine water-quality indicatorsusing MODIS medium-resolution bands: initial results from TampaBay, FL. Remote Sensing of Environment 93, 423–441.

Jayawardena, A.W., Lai, F., 1989. Time series analysis of water qualitydata in Pearl River, China. Journal of Environmental Engineering 115(3), 590–607.

Justesen, P., Ellegaard, A.C., Bernitt, L., Lu, Q.M., 1996. 2-D & 3-Dmodeling of Hong Kong waters. In: Proceedings of the SecondInternational Conference of Hydrodynamics, Balkema, Rotterdam,pp. 649–654.

Lavery, P., Pattiaratchi, C., Wyllie, A., Hick, P., 1993. Water qualitymonitoring in estuarine waters using the Landsat Thematic Mapper.Remote Sensing of Environment 3, 268–280.

Oron, Gideon, Gitelson, Anatoly, 1996. Real-time quality monitoring byremote sensing of contaminated water-bodies: waste stabilization pondeffluent. Water Research 30 (12), 3106–3114.

Ostlund, C., Flink, P., Strombeck, N., Pierson, D., Lindella, T., 2001.Mapping of the water quality of Lake Erken, Sweden, from imagingspectrometry and Landsat Thematic Mapper. The Science of the TotalEnvironment 268, 139–154.

Pattiaratchi, C., Lavery, P., Wyllie, A., Hick, P., 1994. Estimates of waterquality in coastal waters using multi-date Landsat Thematic Mapperdata. International Journal of Remote Sensing 15, 1571–1584.

Tassan, S., 1994. Local algorithms using SeaWiFS data for the retrieval ofphytoplankton, pigments, suspended sediment, and yellow substancesin coastal waters. Applied Optics 33 (12), 2369–2378.

Wang, Y.P., Xia, H., Fu, J.M., Sheng, G.Y., 2004. Water quality changein reservoirs of Shenzhen, China: Detection using LANDSATyTMdata. The Science of the Total Environment 328, 195–206.

Yin, K.D., 2003. Influence of monsoons and oceanographic processes onred tides in Hong Kong water. Marine Ecology Progress Series 262,27–41.

Zhao, H.T., 1990. The Evolution of the Zhujiang River Estuary. ChinaOcean Press, Beijing, China.