the spectral responses of a submerged plant vallisneria spiralis with varying biomass using...
TRANSCRIPT
PRIMARY RESEARCH PAPER
The spectral responses of a submerged plant Vallisneriaspiralis with varying biomass using spectroradiometer
Lin Yuan Æ Li-Quan Zhang
Received: 4 January 2006 / Revised: 21 October 2006 / Accepted: 28 October 2006 /Published online: 1 February 2007� Springer Science+Business Media B.V. 2007
Abstract The relationship between land features
and their spectral characteristics is a key for the
interpretation of remote sensing images. This study
was designed to investigate the spectral responses
of Vallisneria spiralis, a common submerged
aquatic plant in Shanghai, with varying biomass
both in the laboratory and in the Middle Lake
section of a field-scale constructed wetland, using a
FieldSpecTM Pro JR Field Portable Spectroradi-
ometer. The results showed that the reflectance
rate of V. spiralis increased with its increasing
biomass, and this was exhibited both at the visible
band (500–650 nm) and the near infrared band
(700–900 nm). The water environment influenced
the reflectance rate and the primary differences
between the laboratory and field results mainly
occurred at the near-infrared band (700–900 nm).
A regression analysis was carried out between the
biomass of V. spiralis and the reflectance rate at the
wavelengths of QuickBirdTM bands where the
biomass responded most strongly. The results of
this analysis showed a clear linear relationship by
which the biomass of V. spiralis could be quanti-
tatively deduced from the reflectance rate mea-
sured in situ. The implications of this observation,
in terms of the ability of hyperspectral remote
sensing to estimate and monitor the distribution
and dynamics of submerged aquatic vegetation on
a large scale, are discussed.
Keywords Submerged plant �Vallisneria spiralis �Biomass � Reflectance rate � spectroradiometer
Introduction
Submerged aquatic vegetation (SAV) is an
important component of aquatic ecosystems, as
it provides food and shelter for wildlife, and
habitat for spawning aquatic animals. SAV also
retain nitrogen and phosphorus, remove excess
nutrients and reduce the growth of algae (Jin,
2001). The reestablishment of submerged aquatic
vegetation has been recognized as a key measure
for restoring eutrophied lakes or rivers (Qiu et al.,
1997; Pu et al., 2001; Zhong et al., 2003).
Vallisneria spiralis is a completely submerged
plant commonly occurring in the lakes, ponds and
streams of the drainage area along the middle and
lower reaches of the Yangtze River, China. This
plant is now widely used as an ecological engi-
neering species for aquatic ecosystem restoration
in this region (Jin, 2001).
Handling editor: S. M. Thomaz
L. Yuan � L.-Q. Zhang (&)State Key Laboratory of Estuarine and CoastalResearch, Shanghai Key Laboratory for Urbanizationand Ecological Restoration, East China NormalUniversity, 3663 Zhongshan Road North,Shanghai 200062, Chinae-mail: [email protected]
123
Hydrobiologia (2007) 579:291–299
DOI 10.1007/s10750-006-0444-1
Mapping the abundance and distribution of
aquatic biomass is essential to the management of
SAV. However, to map the distribution and
monitor the growth and dynamics of SAV on a
large scale is very labor intensive and time-
consuming due to the restriction of the water
environment. In the recent years, remote sensing
has become an important tool for vegetation
description and classification on a large scale
(Roughgarden et al., 1991). Unlike remote sensing
of terrestrial vegetation, however, the radiation
reflected from SAV must cross the air–water
interface. In addition, certain optically active
components in the water column, such as algal
chlorophyll and suspended contents, may alter the
spectral signal of the SAV. A number of research-
ers have investigated the potential of using remote
sensing approaches to provide timely data for the
mapping and monitoring of SAV (Orth & Moore,
1983; Welch et al., 1988; Nohara, 1991; Jensen,
1992; Jensen et al., 1993; Kirkman, 1996; Jakub-
auskas et al., 2000; Han & Rundquist, 2003;
Williams et al., 2003). Roy (1993) used Landsat
Thematic Mapper(TM) to assess seagrass biomass
in the southern Exuma Cays, and the spectral
characteristics of V. spiralis with varying coverage
were measured with a ground sensor/radiometer
(Yuan & Zhang, 2006), but the quantitative rela-
tionship between the biomass of SAV and its
spectral reflectance has yet to be fully explored.
The goal of this research was to investigate the
spectral characteristics of SAV with a view to
determining the utility of remotely sensed data in
mapping and monitoring its distribution and for
quantitative estimation. To address this goal, the
spectral responses of a major submerged plant (V.
spiralis) with varying biomass, both in the laboratory
and at a constructed wetland in Shanghai were
examined. A number of questions were considered.
First, does this submerged plant have its own unique
spectral signature? Second, can the biomass be
quantitatively estimated by the spectral reflectance
measured? Finally, what are the most important
factors influencing the spectral characteristics of
SAV? Identifying the relationship between the
biomass of SAV and its spectral characteristics is an
important first step to providing timely data for the
estimating, monitoring and managing of aquatic
biological productivity on a large scale.
Materials and methods
Materials
V. spiralis is a dominant perennial submerged
aquatic plant commonly occurring in the lakes,
ponds and streams of the Shanghai region. The
plant is an herbaceous and fast-growing species
with narrow, long leaves based on the stolons
procumbent in the sediments of the water body
(Yan, 1983). This species can remove excess
nutrients in the water body and has been widely
used as an appropriate ecological engineering
species for aquatic ecosystem restoration in this
region. V. spiralis was thus selected as the
experimental material in this study.
Laboratory experiments
A series of plastic boxes measuring 32 · 26 ·23 cm each, filled with 5 cm clear sands on the
bottom and 15 l of 10% Hoagland culture fluid
were prepared for the transplantation. Normal
and healthy individuals (each ca. 15 cm long and
with 3–4 leaves) of V. spiralis were transplanted
into each box, with coverage varying between 0%
and 80% in May 2005. Because the plants we
transplanted had similar heights, each box repre-
sented a varying biomass of V. spiralis. These
boxes were then transferred into growth cham-
bers with a 16/8 h light/dark regime
(9000 l · light intensity) at a constant tempera-
ture of 26�C for 40 days.
Field experiments
The field experiments were conducted out-doors
in natural sunlight at the Middle Lake (625 m2
surface area and average depth 90 cm) of a
constructed wetland in Mengqing Garden on the
southern bank of Suzhou Creek (30�23¢ N and
120�50¢ E), Shanghai. A series of plastic boxes
measuring 45 · 28 · 23 cm each, were filled with
5 cm of lake sediments. Normal and healthy
individuals of V. spiralis (each ca. 15 cm long
and with 3–4 leaves) were transplanted into each
box, with coverage varying between 0% and 80%,
this represented the varying biomass at the field
site in March 2005. The boxes were attached with
292 Hydrobiologia (2007) 579:291–299
123
nylon ropes and were lowered into the lake, thus
allowing the experimental plants to grow in a
natural condition.
Spectral data collection
The spectral measurements for each box with
varying biomass of V. spiralis were made in June
2005 using a FieldSpecTM Pro JR Field Portable
Spectroradiometer produced by ASD Inc. The
instrument collects data over the range
350~2,500 nm, with a sampling interval of
1.4 nm between 350–1,000 nm and 2 nm for the
region between 1,000 and 2,500 nm. The sensor
head used with the instrument had a full angle
field of view of 25 degrees. All measurements
were made from a nadir position, with the sensor
head located 0.5 m above the water surface,
giving an approximate sensor footprint of
0.22 m. The canopy of each box with varying
biomass of V. spiralis was always adjusted to be
approximately 10 cm below the water surface.
Twenty measurements were made for each box
and optimization of the ASD was carried out
approximately every 20 min, using a spectrolon
panel of known reflectance (20%).
In the laboratory experiment, the walls of the
box were covered with a black cotton cloth to
eliminate extraneous internal reflectance and a
50-W Tungsten-Quartz-Halogen lamp was used
as a source of artificial illumination. In the field
experiment, the measurements were made be-
tween 10:30 h and 14:30 h local time under
uniformly cloudless sunny weather conditions.
The concentration of chlorophyll and suspended
contents in the water body were 10.5 lg/l and
12.0 mg/l respectively during the spectral mea-
surements. At the same time as spectral measure-
ments were taken, the fresh weight (FW),
coverage and average height of V. spiralis for
each box were measured both in the laboratory
and in the field experiments. The biomass of
V. spiralis for each box was normalized to
g FW/m2 (Table 1).
Spectral data processing
The spectral data were processed to reflectance
values using the ViewSpecTM Pro 4.02 software
provided by ASD Inc. Twenty reflectance values
for each box with varying biomass of V. spiralis
were then averaged in order to eliminate any
potential influence of variations in illumination.
Due to an observed low signal-to-noise ratio at
wavelengths shorter than 400 nm and longer than
900 nm, only reflectance calculated between 400
and 900 nm was used. Since the study intended to
look at the potential application of airborne
sensors for biomass estimation, the averaged
reflectance values for each box were processed
to represent the four bands of QuickBirdTM, i.e.
blue light band (450–520 nm), green light band
(520–600 nm), red light band (630–690 nm) and
near infrared light band (760–900 nm).
Identifying the relationship between the vary-
ing biomass and its spectral characteristics is
essential in addressing the questions stated in the
introduction. The approach adopted here was
statistically based, using correlation and regres-
sion to analyze their quantitative relationship.
Table 1 The biomass, coverage and average height ofVallisneria spiralis in the laboratory and field experiments.B1–B6 and B1¢–B6¢ are the sample number with varying
biomass of V. spiralis measured in the laboratory exper-iment and field experiment, respectively
Laboratory Field
Sample No. Biomass(g/m2)
Coverage (%) Averageheight (cm)
Sample No Biomass(g/m2)
Coverage (%) Averageheight (cm)
B1 0 0 – B1¢ 0 0 –B2 250.68 20 19 B2¢ 257.68 20 23B3 496.08 30 20 B3¢ 518.48 30 28B4 926.76 50 17 B4¢ 944.08 50 20B5 1441.44 70 22 B5¢ 1497.76 70 23B6 2010.84 80 20 B6¢ 2054.56 80 27
Hydrobiologia (2007) 579:291–299 293
123
The correlation and regression analyses were
implemented using SPSS11.0 and OriginLabTM
7.5 packages.
Results
Laboratory experiments
The spectral reflectance curves for the varying
biomass of V. spiralis in the laboratory experi-
ments formed a typical vegetation spectral curve,
with a small peak around 550 nm (the ‘green
peak’), higher reflectance beyond 700 nm (near-
infrared; NIR) than in the visible spectrum and
two low intensity absorption areas in the blue
band (between 400 and 500 nm) and red band
(around 675 nm) (Fig. 1). Figure 1 also depicts
the reflectance responses to the varying biomass
of V. spiralis, where it is seen that the reflectance
decreased in an orderly fashion with decreasing
biomass, especially in the NIR range, i.e. from
about 9% of reflectance with high biomass to 6%
of reflectance with low biomass; the pattern was
less obvious below 700 nm. It is also noticed that
the reflectance curves for the 0g biomass treat-
ment (100% water without any submerged plant)
did not show a typical vegetation spectral curve,
although it also had higher reflectance in the NIR
band. Therefore, the reflectance response of V.
spiralis with varying biomass was mainly observed
around 575 nm and in the NIR.
To understand the strength of the association
between the biomass of V. spiralis and the
spectral reflectance, the Pearson product-mo-
ment correlation coefficient r was estimated. A
strong positive correlation (P < 0.05) was found
at roughly the wavelength range of 400–421 nm
and 722–900 nm, with the highest correlation
coefficient around 887 nm (r = 0.967, P < 0.01)
(Fig. 2). This indicated that the higher the
biomass of V. spiralis, the stronger the reflec-
tance was at those wavelengths. Therefore,
regression analysis between the biomass of V.
spiralis and the reflectance at 887 nm and at the
band 4 (760–900 nm) of QuickBirdTM were
performed. A clear pattern of increasing reflec-
tance across an increase in biomass from left to
right can be identified (Fig. 3). The Pearson
product-moment correlation between the vary-
ing biomass of V. spiralis and its reflectance at
887 nm was highly significant (R2 = 0.9278;
P < 0.01) (Fig. 3a). The correlation between
the varying biomass of V. spiralis and its
reflectance at the band 4 of QuickBirdTM was
also highly significant (R2 = 0.9009; P < 0.05)
(Fig. 3b).
Field experiments
A similar pattern was observed in the field exper-
iments, i.e. as the biomass of V. spiralis increased,
the reflectance increased around 550 nm and
beyond 700 nm, and formed absorption areas
400 500 600 700 800 9000
2
4
6
8
10
Ref
lect
ance
(%
)
Wavelength (nm)
B1
B2
B3
B4
B5
B6
Fig. 1 Reflectance of Vallisneria spiralis with varyingbiomass in the laboratory experiments (the legends ofB1–B6 see Table 1)
400 500 600 700 800 900
-0.4
0.0
0.4
0.8
1.2
r a=0.05 a=0.01
Co
rrel
atio
n c
oef
fici
ent
( r)
Wavelength (nm)
Fig. 2 Correlation coefficients between the Vallisneriaspiralis of varying biomass and their reflectance measuredin the laboratory
294 Hydrobiologia (2007) 579:291–299
123
between 400–500 nm and around 675 nm (Fig. 4).
However, the reflectance maximum around
550 nm (the ‘green peak’) were much higher than
that in the clear water of the laboratory experi-
ments and the ‘NIR plateau’ evolved into two
peaks with one around 720 nm and the other
around 830 nm.
Figure 5 depicts the Pearson product-moment
correlation coefficient r between the biomass of V.
spiralis and spectral reflectance in the field exper-
iments. A strong positive correlation with biomass
was found at the wavelength between 436–700 nm
(P < 0.01) and 835–900 nm (P < 0.05), with the
highest correlations at 533 nm (r = 0.986) in the
visible band and at 895 nm (r = 0.878) in the NIR.
A clear pattern of increasing reflectance with an
increase in biomass from left to right could be
identified at 533 nm (R2 = 0.9391, P < 0.01) and
895 nm (R2 = 0.741) (Fig. 6a, b).
To explore the potential of using current
operational satellite sensors for the estimation
and monitoring of SAV, regression analyses
were performed between the varying biomass of
V. spiralis and the four bands of QuickBirdTM.
A clear linear relationship was found for the
blue light band (450–520 nm, R2 = 0.9696,
P < 0.01), the green light band (520–600 nm,
R2 = 0.9605, P < 0.01), the red light band (630–
690 nm, R2 = 0.986, P < 0.01) and the NIR band
(760–900 nm, R2 = 0.8186, P < 0.05)(Fig. 6c–f).
y = 0.0016x + 3.6361
R2 = 0.9278
2
3
4
5
6
7
8
-200 300 800 1300 1800 2300
Biomass (g/m2)
Ref
lect
ance
(%
)y = 0.0014x + 4.7543
R2 = 0.9009
4
5
6
7
8
-200 300 800 1300 1800 2300
Biomass (g/m2)
(a) (b)
Fig. 3 Regression analysis between the biomass of Vallisneria spiralis and their reflectance measured in the laboratory: (a)at 830 nm and (b) at the Quick Bird band 4 (760–900 nm)
400 500 600 700 800 900
2
3
4
5
6
7
Ref
lect
ance
(%
)
Wavelength (nm)
B1' B
2' B
3'
B4' B
5' B
6'
Fig. 4 Reflectance of Vallisneria spiralis with varyingbiomass in the field experiments (the legends of B1
¢ –B6¢
see Table 1)
400 500 600 700 800 9000.5
0.6
0.7
0.8
0.9
1.0
r a=0.05 a=0.01
Co
rrel
atio
n c
oef
fici
ent
(r)
Wavelength (nm)
Fig. 5 Correlation coefficients between the Vallisneriaspiralis of varying biomass and their reflectance measuredin the field
Hydrobiologia (2007) 579:291–299 295
123
Discussion
Spectral characteristics of Vallisneria spiralis
with varying biomass
A generally positive correlation between vegeta-
tion biomass, usually measured by leaf area index
(LAI) or coverage, and NIR reflectance has been
observed for common terrestrial vegetation (Gao
& Zhang, 2006). Yuan & Zhang (2006) reported
that the reflectance rate of V. spiralis increased
with its increasing coverage at the visible band
and the NIR band. Zhang (1998) used a Landsat
5TM image to assess the total biomass of SAV in
Honghu Lake, Hubei Province, China, and found
that submerged vegetation biomass had a good
linear relationship with spectral reflectance. Our
results both from the laboratory and field exper-
iments for the SAV plant V. spiralis displayed a
typical vegetation spectral curve, which showed
similar trends to those studies.
The reflectance of SAV in the visible bands
and in the NIR was less than 10%, which
was relatively low compared to the common
y = 0.0006x + 3.5189R2 = 0.9391
2
3
4
5
6
-200 300 800 1300 1800 2300
Ref
lect
ance
(%
)
y = 0.0008x + 2.6224R2 = 0.741
1
2
3
4
5
-200 300 800 1300 1800 2300
y = 0.0058x + 3.0335R2 = 0.9696
2.7
2.9
3.1
3.3
3.5
3.7
-200 300 800 1300 1800 2300
Ref
lect
ance
(%
)
y = 0.0183x + 3.5574R2 = 0.9605
3
3.5
4
4.5
5
5.5
-200 300 800 1300 1800 2300
y =0.0006x + 3.4378R2 = 0.986
3
3.5
4
4.5
5
-200 300 800 1300 1800 2300
Biomass (g/m2)
Ref
lect
ance
(%
)
y = 0.0167x + 3.5614R2 = 0.8186
3
3.5
4
4.5
5
5.5
-200 300 800 1300 1800 2300
Biomass (g/m 2)
(a) (b)
(c) (d)
(e) (f)
Fig. 6 Regression analysis between the biomass of Vallis-neria spiralis and their reflectance measured in the field:(a) at 533 nm, (b) at 895 nm, (c) at the Quick Bird band 1
(450–520 nm), (d) at the Quick Bird band 2 (520–600 nm),(e) at the Quick Bird band 3 (630–690 nm), and (f) at theQuick Bird band 4 (760–900 nm)
296 Hydrobiologia (2007) 579:291–299
123
terrestrial vegetation and could be attributed to
the facts that the radiation reflected from SAV
must cross the air–water interface. However, a
clear linear relationship between varying biomass
of V. spiralis and their spectral reflectance at the
wavelength between 500–650nm and at the NIR
(700–900 nm) could be identified in our research.
Therefore, this approach demonstrated a poten-
tial of measuring quantitatively the distribution
and biomass of SAV by deducing from the
reflectance rate measured in situ.
Comparison between the laboratory and field
experiments
The primary differences in the spectral signature
between the laboratory and field experiments were
the higher reflectance maximum around 550 nm
(the ‘green peak’), the lower ‘red edge’ near
700 nm and the two peak NIR ‘plateau’ in the field
experiments. In general, variation in spectral sig-
nature is determined by factors such as canopy
physiognomy and physiology, the proportion of
green plant material in the canopy, the orientation
of individual leaves, the characteristics and degree
of visibility of the substrate and the amount of
canopy shadow (Armitage et al., 2004). In this case,
the plant material was the same and the structure of
the canopies was very similar, thus the differences
could be attributed to clear water in the laboratory
experiments versus algal-laden water in the field
experiments. The algal chlorophyll and other
suspended contents in the field water could prob-
ably change the reflectance at the visible bands and
the NIR (Han & Rundquist, 2003). Therefore, care
must be taken when the spectral signature of SAV
measured in the laboratory is to be used in the field.
Further work is needed to understand the impacts
of the water environment on the composite spec-
tral signature of SAV.
Implications for current operational satellite
and airborne systems
The restoration and reconstruction of SAV is now
a required action for restoring eutrophied lakes or
rivers in China. However, mapping the distribu-
tion and monitoring the growth and dynamics of
SAV on a large scale is very labor intensive and
time-consuming, e.g. large lakes, difficulties in
accessing sites and getting enough sites for reli-
able data, etc. A number of researchers have
investigated the potential of using current oper-
ational airborne systems for the mapping and
monitoring of SAV. A Landsat 5TM image was
used to assess the total biomass of SAV in
Honghu Lake, by deducing from the relationship
between SAV biomass and its spectral data
measured at a series of sampling sites (Zhang,
1998). In that study, the weighed SAV biomass
was set into a tank filled with the lake water and
the spectral reflectance was measured, in a
manner similar to our laboratory experiments.
As mentioned previously, the structure and shape
of the SAV canopy as well as the water environ-
ment conditions might be different in this mea-
surement, and the ground based in situ spectral
data could greatly improve the overall accuracy of
SAV biomass assessment.
Roy (1993) used Landsat Thematic Map-
per(TM) to assess the seagrass biomass in the
southern Exuma Cays, Bahamas. The TM bands
were transformed to minimize the depth-dependent
variance in the bottom reflectance signal. Williams
et al. (2003) applied airborne hyperspectral remote
sensing imagery for automated mapping of SAV in
the tidal Potomac River, Maryland, USA, by
developing a spectral library database containing
ground-based and airborne sensor spectra. The
depth of the absorption feature at a specific
wavelength could be used to identify two species
of SAV. With the advent of satellite sensors such as
QuickBirdTM and IkonosTM, which have spatial
resolutions similar to airborne systems, work by
Schmidt and Skidmore (2003) has indicated that
using greater spectral resolution in the form of
hyperspectral data may produce better results.
In this research, the ground-based spectral
signatures of V. spiralis with varying biomass
were measured both in the laboratory and field
conditions. The clear pattern of linear relation-
ships between the biomass of V. spiralis and their
spectral reflectance at the wavelengths of Quick-
BirdTM bands will be useful with the current and
forthcoming space-based hyperspectral remote
sensing systems to map SAV distributions and
abundance, and to estimate the biomass of SAV
in a shallow water body.
Hydrobiologia (2007) 579:291–299 297
123
In conclusion, this research has only investi-
gated a single species of SAV and a limited
sample of aquatic environments in Shanghai.
Identifying the relationships between the biomass
of SAV and their spectral characteristics is an
important first step to providing timely data for
the mapping, monitoring, estimating and manag-
ing of SAV on a large scale. Further work on
other types of SAV using multi-seasonal spectral
data, and investigating the impact of the aquatic
environment and water quality on the composite
spectral signature of SAV is necessary in order to
determine whether similar patterns emerge. Fur-
thermore, future work must also consider a
comparative analysis with hyperspectral airborne
remote sensing systems such as CASI, AVIRIS,
PHI and OMIS, or hyperspectral satellite remote
sensing system such as Hyperion for the tested
areas, although these hyperspectral airborne or
satellite data are usually difficult to acquire at the
moment.
Acknowledgements The authors would like to thankmembers of the Ecological Section of the State KeyLaboratory of Estuarine and Coastal Research, East ChinaNormal University, for their assistance with the collectionof the field data. We also thank Professor Bruce Anderson,Queen’s University, Canada for valuable comments andlinguistic checking. The research has been funded by thekey project of the Shanghai Scientific & TechnologicalCommittee (05DZ12009), National Key FundamentalResearch and Development Program (2003AA601020)and the state’s 10th five-year ‘‘211 Project’’ – supportedkey academic discipline program of ECNU.
References
Ackleson, S. G. & V. Klemas, 1987. Remote sensing ofsubmerged aquatic vegetation in Lower ChesapeakeBay: a comparison of Landsat MSS to TM imagery.Remote Sensing of Environment 22: 235–248.
Armitage, R. P., M. Kent & R. E. Weaver, 2004.Identification of the spectral characteristics of Britishsemi-natural upland vegetation using direct ordina-tion: a case study from Dart moor, UK. InternationalJournal of Remote Sensing 25: 3369–3388.
Gao, Z. G. & L. Q. Zhang, 2006. Identification of thespectral characteristics of natural saltmarsh vegetationusing indirect ordination: a case study from Chong-ming Island, Shanghai, China. Acta PhytoecologicaSinica 30: 252–260.
Goodin, D., L. Han, R. N. Fraser, D. C. Rundquist &W. A. Stebbins, 1993. Analysis of suspended solids inwater using remotely sensed high resolution deriva-
tive spectra. Photogrammetric Engineering and Re-mote sensing 59: 505–510.
Han, L. & D. C. Rundquist, 1994. The response of bothsurface reflectance and the underwater light field tovarious levels of suspended sediments: Preliminaryresults. Photogrammetric Engineering and RemoteSensing 60: 1463–1471.
Han, L. & D. C. Rundquist, 2003. The spectral responsesof ceratophyllum demersum at varying depths in anexperimental tank. International Journal of RemoteSensing 24: 859–864.
Jakubauskas, M., K. Kindscher, A. Fraser, D. Debinski &K. P. Price, 2000. Close-range remote sensing ofaquatic macrophyte vegetation cover. InternationalJournal of Remote Sensing 18: 3533–3538.
Jensen, J. R., 1992. Measurement of seasonal and yearlyaquatic macrophyte changes in a reservoir usingmultidate SPOT panchromatic data. Technical Pa-pers, American Society for Photogrammetry andRemote Sensing 1: 167–176.
Jensen, J. R., S. Narumalani, O. Weatherbee & H. E.Mackay, 1993. Measurement of seasonal and yearlycattail and waterlily changes using multidate SPOTpanchromatic data. Photogrammetric Engineeringand Remote Sensing 59: 519–525.
Jin, X. C., 2001 Technologies of lake eutrophicationcontrol and management. Chemistry Industry Press,Beijing.
Kirkman, H., 1996. Baseline and monitoring methods forseagrass meadows. Journal of Environmental Man-agement 47: 191–201.
Nohara, S., 1991. A study on annual changes in surfacecover of floating-leaved plants in a lake using aerialphotography. Vegetatio 97: 125–136.
Orth, R. J. & K. A. Moore, 1983. Chesapeake Bay: anunprecedented decline in submerged aquatic vegeta-tion. Science 222: 51–53.
Pu, P. M., G. X. Wang, Z. K. Li, C. H. Hu, B. J. Chen,X. Y. Cheng, B. Li, S. Z. Zhang & Y. Q. Fan, 2001.The degradation and restoration of healthy aquaticecology—theory, technique and application. Journalof Lake Sciences 13: 193–203.
Qiu, D. R., Z. B. Wu, B. Y. Liu, G. A. Yan & Y. J. Zhou,1997. The research of aquatic plant communityrestoration experiment. Journal of Lake Sciences 9:168–174.
Roughgarden, J., S. Running & P. Matson, 1991. Whatdoes remote sensing do for ecology. Ecology 72: 1981–1922.
Roy, A. A., 1993. Remote sensing of submerged vegeta-tion canopies for biomass estimation. InternationalJournal of Remote Sensing 14: 621–627.
Schmidt, K. S. & A. K. Skidmore, 2003. Spectral discrim-ination of vegetation types in a coastal wetland.Remote Sensing of Environment 85: 92–108.
Welch, R., R. R. Remilliard & R. B. Slack, 1988. Remotesensing and geographic information system tech-niques for aquatic resource evaluation. Photogram-metric Engineering and Remote Sensing 54: 177–185.
Williams, D. J., N. B. Rybicki, A. V. Lombana, T. M.O’Brien & R. B. Gomez, 2003. Preliminary investi-
298 Hydrobiologia (2007) 579:291–299
123
gation of submerged aquatic vegetation mappingusing hyperspectral remote sensing. EnvironmentalMonitoring and Assessment 81: 383–392.
Yan, S. Z., 1983. The scheme of chinese aquatic macro-phytes. Science Press, Beijing.
Yuan, L. & L. Q. Zhang, 2006. Identification of thespectral characteristics of a submerged plant Val-lisneria spiralis. Acta Ecologica Sinica 26: 1005–1011.
Zhang, X. Y., 1998. On the estimation of biomass ofsubmerged vegetation using Landsat thematic mapper(TM) imagery: a case study of the Honghu Lake, PRChina. International Journal of Remote Sensing19(1): 11–20.
Zhong, Y. X., H. Y. Hu & Y. Qian, 2003. Advances inutilization of macrophytes in water pollution control.Techniques and Equipment for Environmental Pollu-tion Control 2: 36–40.
Hydrobiologia (2007) 579:291–299 299
123