(NIR) bands, no light will exit the water in these bands vir-
tually, except the most turbid coastal waters, thus the radi-
ance measured by the sensor originates from the scattering
of solar irradiance by the atmosphere and water surface
(Gordon and Franz, 2008). Therefore, these bands can be
used to estimate the atmospheric effects. However, it has
been demonstrated that this approach leads to considerable
errors over turbid inland or coasta l waters, where sus-
pended sediments or high concentrations of phytoplankton
and detrital particles may cause a non-negligible reflectance
at near-infrared channels (Guanter et al., 2010). An
approach for the atmospheric correction of MODIS data
over coastal waters based on aerosol retrieval from wave-
lengths greater than 860 nm was proposed by Gao et al.
(2007), where the contribution of suspended particles and
bottom scattering was supposed to be at a minimum. Other
approaches perform the simultaneous retrieval of atmo-
spheric and water components by a multi-parameter inver-
sion using the complete visible and near-infrared
information as well as coupled atmospheric and bio-optical
radiative transfer models, so that aerosol and water param-
eters are retrieved consistently (Lavender et al., 2005;
Moore et al., 2009). These models are adequate for the rep-
resentation of coupled water–atmosphere radiative transfer
problem, and able to provide a pixel-wise description of
horizontal variations in the atmosphere and water. How-
ever, these methods are typically site-specific, as the inver-
sion results depend on the input values applied to constrain
the bio-optical mod el (Guanter et al., 2010).
Generally, it was suggested by Chen et al. (2011b) that it
was more likely to successfully find “clear water” from high
spatial resolution data than low spatial resolution data,
which contained more complicated mixed pixels. Further-
more, two synchronous images of Advanced Wide-Field
Sensor (56 m pixels) and Linear Imaging Self-Scanner
(24 m pixels) of Indian remote sensing satellite were used
by Chen et al. (2011b) to reveal that the uncertainty was
9% at NIR band in estimating “clear water” reflect ance
in Case II waters while data spatial resolution changed
from 24 to 56 m. It is well known that a 5% uncertainty
in retrieved water-leaving radiance at blue bands would
result in a 35% uncertainty in the estimated chlorophyll-a
concentration (Hu and Carder, 2002; Moore et al., 2009).
Thus, the 9% uncertainty associated with scale effects on
pseudo “ clear water” would greatly impact the accuracy
of satellite-derived water-leaving.
The MODIS instrument contains thirty-six spectral
bands at three differen t spatial resolutions with nominal
ground fields of 250, 500 m, and 1 km viewing (Nishihama
et al., 1998). Due to smaller observation scale, the “clear
water” pixel extracted from MODIS 250 m resolution data
is more reasonable and reliable than that of MODIS 1 km
resolution data. Therefore, the accuracy in water-leaving
reflectance estimat ion from MODIS 1 km resolution data
can be improved, if the “clear water” pixel extracted from
MODIS 250 m resolution data is used in atmospheric cor-
rection. The main objectives of this study are: (1) to study
the optimal spectral relationships of MODIS (OSRLM)
250 m and 1 km resolution data at NIR bands observed
at the water–surface and top-of-atmosphere (TOA) respec-
tively; (2) to evaluate the performance of “clear water”
atmospheric correction model (CWAC) in estimating
water-leaving reflectance in Taihu Lake; (3) to improve
the CWAC model by extending “clear water” reflectance
of MODIS 250 m data over that of MODIS 1 km data
(Atmospheric correction model for MODIS 1 km
data using “clear water” reflectance of MODIS 250 m data
(ACMM)); (4) to evaluate the performance of ACMM
model in water-leaving reflectance deriving in Case II
waters.
2. Materials and met hods
2.1. Study areas
Located between longitudes 119°54
0
E and 120°36
0
E, and
latitudes 30°56
0
N and 31 °33
0
N, China’s Taihu Lake pro-
vides normal water usage for several million residents in
the adjacent Wuxi City. Thus, water quality in inland fresh-
water lakes such as Lake Taihu is vital to human activities
and living needs as a critical role in the regional ecosystem,
which may also impact climate changes. Taihu Lake is the
third-largest inland freshwater lake in China with average
water depth of 2 m (Xu et al., 2001). Waters in Taihu Lake
are consistently highly turbid, except East Taihu Bay and
some of the East Lake regions with often clear waters
(Wang et al., 2011). In addition, Taihu Lake has frequent
algae bloom pollutions in the spring–summer. The spring
2007 blue–green algae (Microcystis) bloom in Taihu Lake
caught the worlds’ attention. Algae-polluted waters in the
lake have adversely affected and interrupted the normal life
of several million adjacent residents. Thus, there is an
urgent need to effectively monitor and manage water qual-
ity in Taihu Lake as well as be tter understand the optical,
biological and ecological processes and phenomena in this
fresh waters (Wang et al., 2011).
Recent advances in optical sensor technology have
opened new opportunities to study biogeochemical pro-
cesses in aquatic environments at spatial and temporal
scales that were not possible before (Clavano et al., 2007).
These advances allowed scientists to use satellite images
for synoptically investigating large-scale surface features
in Taihu Lake (Chen et al., 2011c, 2010; Yang et al.,
2006). However, the water-leaving reflectance detected by
satellite sensor is fairly weak. For example, the component
of measured reflectance, backscattering out of the water and
transmitting to TOA is less than 10% in the blue band, typ-
ically much smaller in the green and near-infrared bands
(Gordon and Castano, 1989; Gordon and Clark, 1981).
To get useful water- leaving information from Taihu Lake,
atmospheric absorption and scattering effects must be elim-
inated. Thus, it is significant to carry out the atmospheric
correction study in this region, so as to effectively monitor
water color in Taihu Lake.
J. Chen et al. / Advances in Space Research 51 (2013) 1750–1760 1751