Você está na página 1de 7

Chinese Journal of Oceanology and Limnology

Vol. 24 No. 4, P. 428-434, 2006

Retrieval of chlorophyll a and suspended solid concentrations by hyperspectral remote sensing in Taihu Lake,
China*
YANG Dingtian () ,, PAN Delu () , ZHANG Xiaoyu ()
ZHANG Xiaofeng () , HE Xianqiang () , LI Shujing ()

(Key Laboratory of Tropical Marine Environmental Dynamics (LED), South China Sea Institute of Oceanography, Chinese
Academy of Sciences, Guangzhou 510301, China)
(State Key Laboratory of Satellite Ocean Environment Dynamics, Second Institute of Oceanography, SOA, Hangzhou 310012,
China)

Received Apr. 15, 2005; revision accepted Oct. 12, 2005

Abstract
Chlorophyll a (chl-a) and suspended solid concentrations are two frequently used water
quality parameters for monitoring a lake. Traditional measurement of chl-a and suspended solids, requiring laborious laboratory work, which is often expensive and time consuming. Hyperspectral remote-sensing measurement provides a fast and easy tool for estimating water trophic status. In situ hyperspectral data on March 78, July 67, September 20 and December 78, 2004 and the corresponding
water chemical data were used to regress the algorithm of water quality parameters. Results showed that
the peak of water leaving radiance around 700 nm (R700) varied proportionally with chl-a concentration,
and moved to infrared when algal bloom occurred. The reflectance ratio of R702/R685 was well correlated
with chl-a when water surface in no algal bloom case and the correlation coefficient was better if absorption of phycocyanin was considered. The reflectance ratio R620/R531 was highly correlated to the concentration of suspended solids. The relationship between suspended solids and other band groups were also
compared. Secchi disk depth could be calculated by non-linear correlation with suspended solids concentration.
Key words: hyperspectral remote sensing; chlorophyll a; suspended solids; Taihu Lake

1 INTRODUCTION
Chlorophyll a (chl-a) and suspended solids are
two major indexes for water quality evaluation.
Traditional measurement of them is time consuming
and laborious, sampling is often not enough to cover
entire water body. Fortunately, remote sensing (RS)
can offer attentively synoptic, repetitive, consistent,
cost effective and comprehensive spatial and temporal views; however, successful application is limited in lakes having relatively high non-algal turbidity levels. For high concentrations of suspended
solid and chromatic dissolved organic matter, no
well-succeeded procedure has been reported applying RS techniques to measure absorption and reflection of lake water as well as to retrieve the concen-

trations.
Hyperspectral RS has been used widely in recent
years for its continuous waveband being very informative of water chemical feature. Variation in
water constituents significantly affect spectral characteristics significantly, including absorption by
chl-a, dissolved organic matter, and carotenoids,
scattering by particular matter (Kirk, 1994). Spectral
information is sensitive near 440 nm, which is important for having minimum reflectance caused by
chl-a absorption. Conversion of the information can
* Supported by National Natural Science Foundation of China (No.
40576078), Natural Science Foundation of Guangdong Province (No.
5003685), Post-Doctor Foundation of China, Post-doctor Foundation of
Zhejiang Province, Post-Doctor Foundation of Shanghai and the National High-Tech R&D of China (863 Program) (No. 2002AA639490)
** Corresponding author: dtdang@163.net. Tel: +86-20-89023167

No.4

YANG et al.: Hyperspectral remote sensing of chlorophyll a and suspended solid concentrations

help us to recognize certain constituent in water for


trophic state evaluation. In practice, reflectance (R)
ratio of 440 nm to 550 nm (R440/R550) was frequently
used to estimate chl-a concentration in ocean water
(Gordon and Morel, 1983). In inland water, the
minimum near 440 nm is often indistinct in reflectance spectra due to strong absorption by dissolved
organic matter and suspended particulates. Near 490
nm another trough on reflectance curve is shown
due to largely carotenoids absorption (Gitelson et al.
1995; Yacobi et al. 1995). The peak in green range
between 550570 nm was the minimum algal pigments absorption, non-organic suspended matter and
phytoplankton cell walls. The trough near 625 nm
was resulted from absorption by blue algae, having
phycocyanin pigments. Again, the trough at
670680 nm was the outcome of the absorption by
Chl-a. The distinctive peak in red to near infrared
boundary near 700 nm was regarded as light diffusion by chl-a (Gitelson et al., 2000).
In a inland water, hyperspectral RS on chl-a depends mainly on strong absorption wavebands of
chl-a in red section and the diffusion peak near infra-red. Gitelson invented algorithms by using ratio
R700/R675, assuming to depend on chl-a concentration only. On the basis of Gitelsons work, Shu et al
(1998) developed an equation by considering the
absorbance of phycocyanin at 624 nm, using R600/624
or R560/624 for water quality correction, which
yielded a good result. Gons (1999) studied the ratio
of upwelling irradiance to downwelling irradiance
underwater just at 675 nm and 702 nm and backscattering at 778 nm to calculate chl-a in Taihu
Lake (30553133N; 1195412036) and Chaohu
Lake (31253142N; 1171811750E) with
modified equation. Li et al (2002) studied the relationship between chl-a concentration and the reflectance ratio R702/R672 in Chaohu Lake using an experimental equation for hyperspectral retrieval of
chl-a concentration.
Hyperspectral detection of suspended solids developed quickly in recent years. In 1970s and 1980s,
Gordon et al (1978, 1983) developed a set of equations for the water leaving reflectance due to scatter
of suspended particles; however, the equations cannot be used in inland waters because of strong
backscattering in near-infrared wavelengths. Based
on the linear response from water leaving radiance
to satellite-detected reflectance spectra, Li (2000)
developed a model by removing the fluctuating atmospheric effects, showing a good effect on reflecting suspended solids concentration. Han et al.

429

(2003) developed a linear relationship with experiments in a tank, by regrouping different bands, having a sound result too.
These studies focused on finding relationship
between chl-a, suspended solids concentration and
spectral band ratio, which provided us a solution to
the information for retrieval of chl-a and suspended
solids in Taihu Lake. Being practiced in this work,
the aim of this study was to explore the precise and
efficiency of water quality monitoring, serving for
weekly or monthly alert on environmental control.

2 METHODS
RS reflectance measurements were made using
FieldSpec,a special analytical spectral device, made
in USA. To operate it on water, the input lens of
hyperspectral spectroradiometer was held off boat
with a long pole, in a direction 135 off the sun.
Spectral bandwidth of 1 nm in the range of 330860
nm was measured. Reference plate was used to clear
off the effect of air diffusion on reflectance. The
light above the reference plate was measured directly under the sun and sunlight was shaded (for
calculating reflectance of sky light). Water leaving
radiance than was calculated by total reflectance received by lens minus air diffusive reflectance, ie:
Ll = Lt rL s

(1)

where Ll is water leaving radiance, Lt is the total


radiance received by instrument, r is the reflectance
of sky light, can be measured by r= Lt (800~900nm)
/ Ls (800900 nm), the range of r is 2.1%5%, L s
is sky radiance.
RS reflectance was calculated using the following equation:

R = ( S w rS s ) Pp / S p

(2)

where S w , S s , S p is digital number of spectroradiometer towards water, air and reference plate respectively. Pp is reflectance of reference plate. Hyperspectral data was measured in Taihu Lake on
March 78, July 67, September 20 and December
78, 2004, and simultaneously analyzed the water
column for chl-a, suspended solids concentration.
Secchi depth was measured in situ with secchi
disk.
Total suspended solids was measured from 500
ml water samples filtering through a pre-weighted
glass fiber filter.

430

CHIN. J. OCEANOL. LIMNOL., 24(4), 2006

Chl-a concentration was measured fluorometrically using Turner-Design Model 10 Fluorometer


calibrated with pure chl-a (Sigma Chemical Co.).

Vol.24

Hydrographic and water chemical parameters of


the sample are shown in Table 1.

Table 1 Hydrographic and water chemical parameters of the samples*


Water temperature

Wind speed

Chl-a

Dissolved oxygen

Secchi depth

Algal bloom

()

(m/s)

(mg/m3)

(mg/L)

(cm)

occurrence

March 78

7.39.2

0.85.8

7.327.6

10.5511.45

1529

None

July 67

2729.6

0.45.8

4.1172

6.078.21

2158

Most of sites

September 20

22.823.8

N/A

10.160.8

7.498.54

1635

Most of sites

December 78

7.99.4

N/A

5.069.9

7.909.40

13172

Some sites

Date of sampling

* Unites of all indices in the paper are referred to the Table 1, N/A means not available.

Fig.1 Surface reflectance in Taihu Lake, showing variation of peak around 700 nm
a. Case of March 78, 2004, showing the sharp trough at 687 nm; b. Case of July 67, 2004, showing the absorption trough
around 675 nm; c. Case of September 20, 2004; d. Case of December 78, 2004

No.4

431

YANG et al.: Hyperspectral remote sensing of chlorophyll a and suspended solid concentrations

3 RESULTS AND DISCUSSION


3.1 Experimental retrieval of chl-a
Sixty seven ship-borne spectra were acquired on
78, March, July 67, September 20 and December
78, 2004 for retrieval of water quality parameters.
On March 78, strong absorption at 657nm and

687nm resulted in two apparent troughs on the


spectra; strong reflectance around 650 nm and 700
nm caused two peaks on the spectra. While on the
other days of observation (Fig.1a), the trough at 657
nm still remained, but trough at 687 nm moved to
672 nm and became flattened. Low chl-a concentration and no algal bloom on the water surface in
March were probably the reason of trough-and-peak

Fig.2 Relationship between chl-a and R702/R675 in different seasons


a. On March 78, 2004, good linear relationship; b. On July 67, 2004, moderate relationship; c. On September 20, 2004,
scattered due to limited data; d. On December 78, 2004, good linear relationship

Table 2 Relationship between chl-a and R702/R675 in different seasons*


Season

Equation

March 78

y= 212.7x181.1

0.93

2.0

14

<0.0001

July 67

y= 97.22x61.99

0.84

21.3

17

<0.0001

September 20

SD

= 11.93x+3.36

0.81

5.5

0.05003

December 78

y= 58.67x41.17

0.77

8.6

24

<0.0001

Total

y=58.38x35.57

0.65

19.7

61

<0.0001

* y= chl-a concentration; x=R702/R675

432

CHIN. J. OCEANOL. LIMNOL., 24(4), 2006

While high concentration of chl-a and algal bloom


occurrence moved the peak around 700 nm to infrared direction (Fig.1b, c, d). The absorption by chl-a
at 675 nm and the reflectance at 702 nm can be used
for retrieval of chl-a concentration in lake water
(Fig.2). Gitelson et al (1995) proved that the linear
relationship between chl-a concentration and
R702/R675 are applicable in chl-a concentration range
from 180 to 200 mg/m3, but the relation would be
exponential in higher chl-a concentration, which is

Vol.24

also true in Taihu Lake and Chaohu Lake (Shu et al,


2000; Li et al, 2002). Considering the absorption of
phycocyanin (Shu et al, 1998) would yield better
outcome. Relationship between chl-a and spectral
band in Taihu Lake in different seasons was shown
to be linear in regression models (Fig.2), correlation
coefficient was more than 0.65 on average, and
greater than 0.9 in March (Table 2), which may be
explained by low chl-a concentration of below 200
mg/m3.

Table 3 Comparison of the relative coefficient (R2) for retrieval of chl-a concentration in three different equations
Sampling date

Relative coefficient (R2)


March 78

July 67

September 20

December 78

Chl-a = a+b(R702/R675)

0.872 9

0.712 6

0.658 3

0.594 9

Chl-a = a+b1(R702/R675)+b2(R600/R624)

0.880 3

0.723 1

0.690 3

0.874 1

Chl-a = a +b1(dR590)+b2(dR605)+b3(dR610)

0.290 6

0.264 5

0.251 7

0.307 8

a: intercept; b,b1,b2 and b3: slope; d: second derivative

Fig 3 Relationship of suspended solids concentration with other light indicators


a. with R620/R531; b. With band group (R560-R520)/(R560+R520); c. With band group (R550+ R670)/(R490/R550); d. With Secchi disk depth

No.4

YANG et al.: Hyperspectral remote sensing of chlorophyll a and suspended solid concentrations

However, problems were still unsolved. The coefficient varied largely in different equations. In Table 2, in the equation form of y=a+bx, a varies from
-181.1 to +3.36, and b from 11.93 to 212.7. The coefficient is very local and seasonal, which greatly
impacted the accurate retrieval of chl-a concentration. How to deal with the variation in coefficient is
next step of this research.
The relationship between chl-a concentration
and spectra was compared among three different
equations by using observational data for TaiHu
Lake in different seasons. The correlation coefficient R2 can better describe the relation if considering the absorption of phycocyanin (main absorption
peak of phycocyanin is at 624 nm, Table 3).
Ratio of red to near infrared, had been proved
successful for retrieval chl-a in inland water; however the imperfect is that parameters varied greatly
in different water, which is due to: (1) Chlorophyll
specific absorption varied with chl-a concentration,
package effect, pigment composition and photoadaptation; (2) The precondition for using two-band
equation is that the absorption by water and chl-a at
670 nm, scattering by chlorophyll, water and suspended solids at 702 nm, and the scattering of suspended solids being unchanged. In fact this assumption is a little away from the real result; (3) Sensor
noise and light direction affected measurement; (4)
Scattering coefficient of different sizes of suspended
solids varied greatly; (5) Scattering signals at 702
nm tend to be saturated when density of water
bloom exceeded certain concentration; (6) Whitecap
foam etc.
3.2 Retrieval of suspended solids and Secchi disk
depth
Suspended solids are the main component of
water in Taihu Lake, and most of them were inorganic when wind was strong. Reflectance of suspended solids showed two general peaks: the first
peak was at 580680 nm, the second peak was at
760820 nm. The first peak was usually greater than
the second one when the suspended solids concentration is low. When suspended solids exceeded certain concentration, the peak at 760820 nm increased with increasing concentration of suspended
solids (Han et al., 2003).
Hyperspectral data from Taihu Lake in 2004
were also used for retrieval the suspended solids
concentration. Spectra band ratio of R620/R531, spec-

433

tra band groups of (R560-R520)/(R560+R520) and


(R550+R670)/(R490/R550) were used to regress the relationship between spectra and suspended solids. Figure 3a shows the relationship between spectra band
ratio of R620/R531 and suspended solids in March,
having a good linear relationship with correlation
coefficient (R2 ) up to 0.558. Fig 3b shows the correlation between band group (R560-R520)/(R560+R520)
and suspended solid. With the increasing concentration of suspended solids, the value of (R560-R520)/
(R560+R520) increased first and then decreased. R2
can be up to 0.79. Non-linear relationship existed
between suspended solid and spectra band group
(R550+R670)/(R490/R550) with R2 at 0.73 (Fig.3c).
Among the above results, the best relationship between spectra and suspended solids can be described
by spectra band group (R560-R520)/(R560+R520).
Secchi disc depth is largely influenced by suspended solids, both inorganic and organic (Fig. 3d).
So secchi disc depth can be retrieved from suspended solids concentration in exponential relationship:
SDD = 0.27+1.18e(-SS/14.16)
(R2 = 0.70604; n = 657)

(3)

where SDD represents the Secchi disc depth; SS is


the suspended solid.
With the Eq.(3), the Secchi depth can be estimated from the suspended solid, which is also linearly correlated with the reflectance ratio, R620/R531 in
the relationship given below:
SDD=0.27+1.18e(0.8722+0.001(R

620

/R531))/14.16)

(4)

4 CONCLUSION
Application of hyperspectral method for retrieval of chl-a concentration and suspended solid
concentration was successfully applied in Taihu
Lake for facilitating the water quality monitoring
and controlling. Relationship of chl-a and band ratio
of R702/R685 is the best in March, maybe it was benefited with no algal bloom on the water surface.
Chl-a concentration can be well retrieved if the absorption by phycocyanin is considered; however, the
second derivative is not suitable for retrieval of
chl-a concentration in Taihu Lake. Suspended solids
can be well calculated by the band ratio of R620/R531,
and Secchi disk depth can be retrieved also by sus-

434

CHIN. J. OCEANOL. LIMNOL., 24(4), 2006

pended solid. The problem is that the parameters


varied greatly in different seasons and locations. So
we should do it carefully to avoid possible bias and
develop new methods to improve the equation with
better result.
References
Gitelson, A., S. Laorawat, G. P. Keydan et al., 1995. Optical
properties of dense algal cultures outdoors and its application to remote estimation of biomass and pigment
concentration in Spirulina platensis. J. Phycology 31(5):
828-834.
Gitelson.A., Y. Z. Yacobi, D. C Rundquist, J. F. Schalles, R.
Stark, L. Han and D. Etzion, 2000. Remote estimation
of chlorophyll concentration in productive waters: Principals, algorithm development and validation. URL:
http://www.nwqmc.org/2000proceeding/papers/pap_git
elson.pdf
Gons, H. J., 1999. Optical teledetection of chlorophyll-a in
turbid inlands Waters. Environmental Science and
Technology 33:1 127-1 132.
Gordon, H. R and A. Y. Morel, 1983. Remote assessment of
ocean color for interpretation of satellite visible imagery:
a review. Lecture Notes on Coastal and Estuarine Studies. New York: Springer-Verlag. P. 114
Gordon, H. R., 1978. Remove of atmospheric effects from
satellite imagery of the oceans. Applied Optics 17(10):
1 631.

Vol.24

Han Z, C. X. Yun and X. Z. Jiang, 2003. Experimental study


on reflected spectrum of suspended sediment. Journal
of Hydraulic Engineering 12: 117-121. (in Chinese)
Kirk, J. T. O., 1994. Light and Photosynthesis in Aquatic
Ecosystems. Cambridge University Press, Britain, p.
509.
Li, S. J., Q.Wu, X. J.Wang et al., 2002. Correlations between
reflectance spectra and concentrations of chlorophyll a
in ChaoHu lake. The Journal of Lake Science 14:
228-235. (in Chinese)
Li, Y. and J. Li, 2000. A suspended sediment satellite sensing algorithm based on gradient transiting from water-leaving to satellite-detected reflectance spectrum.
Chinese Science Bulletin 45 (10): 925-931.
Shu, X. Z. and D.B. Kuang, 1998. New algorithm to estimate
chlorophyll-a concentration from the spectral reflectance of inland water. SPIE 3 5020277-786X198:
254-258.
Shu, X. Z., Q. Ying and D. B. Kuang, 2000. The relationship
between algal chlorophyll concentration and spectral reflectance of inland water. Journal of Remote Sensing (in
Chinese) 4: 41-45
Yacobi, Y. Z., A.Gitelson and M. Mayo, 1995. Remote
sensing of chlorophyll a in Lake Kinneret using high
spectral resolution radiometer and Landsat TM: Spectral
features of reflectance and algorithm development. J.
Plankton Res. 17: 2155-2173.

Você também pode gostar