IEEE Proof
Web Versi
CHEN et al.: SATELLITE REMOTE SENSING OF CHLOROPHYLL-A CONCENTRATION IN COASTAL WATERS 3
These achievements had a great impact on our knowledge of
ocean and the application of ocean color remote sensing was
dramatically increased in the early 1990s. Owin g to the im -
mature techn olo gies of sensor calibration in orbit, rad iometric
calibration, atmospheric correction, the accuracy of
con-
centration estimation was still n ot very high. Moreover, it was
difficult to m eet the accuracy requirements in som e application
fields, such as the role of oceanic photosynthesis in coastal
waters. However, the achievement and experience were of great
significance to the water color remote sensing and laid a solid
foundation for future researches.
The ocean color sensors related technology had taken its orig-
inal shape but was still immature during the period from the late
1990s to late 2000s. Since the launch of the Sea-viewing Wide
Field-of-view Sensor (SeaWiFS) on September 18, 1997, glob al
ocean color data had been available in near-real time to the
science community [26]. SeaWiFS was on a sun-synchronous
satellite with a 2801 km swath width, providing 2-day coverage
of the global o cean with a nadir resolution of
per pixel
[27]. These data were critical for und erstan din g the temporal
variability of marine ecosystems, especially with regard to
events such as El Niqo and La N iqa, and the role of oceanic
photosynthesis and primary productivity in th e Earth’s carbon
budget and climate [28]. The successful launching of SeaWiFS
had an epoch-marking significance in marine science. Until
then [29], the SeaWiFS provided water-leaving reflectance with
uncertainty and concentration with 35% uncertainty
in Case I waters on a global scale. Shortly afterwards, tw o
Moderate Resolution Imaging Spectroradiometers (MODIS),
including Terra MODIS (since 1999) and Aqua MODIS (since
2002), were launched by NASA. These satellite sensors pro-
vided near-daily coverage of the global ocean. MO DIS was
typically 2–3 times m ore sensitive than SeaWiFS, which in turn
was approximately twice as sensitiv e as CZCS [13]. Analysis
suggested that the pigmen t concentration could be derived from
the radiance ratio of MODIS with an uncertainty of
[30]
in Case I w aters, a d ecrease of
of uncertainty compared
with SeaWiFS. With the later launching o f Nation al Polar-Or-
biting Op e ration al Environm ental Satellite System (NPOESS)
which carried the Visible/Infrared Imager/Radiometer Suite
(VIIRS) instruments, continuity in coverage was expected to
achieve throug h the NPO E SS Preparatory Mission (NPP). In
the last few years, China, Canada, India, and several other
countries have launched satellites providing marine science
data, besides, joint Sino-Euro dragon program has been very
successful [31]. Up to now, a large am ount of satellite data has
been available for remote sensing of
concentration.
However, it is still difficult to fin d out the satellite data with
optimal spectral and spatial resolution for
concentra-
tion estim ation in coastal regions, and this issue has aroused
people’s increasing attention. Even though some oceanic
observation satellite sensors, such as SeaWiFS and MODIS,
have been designed to measure water-leaving radiance related
to
concentration, these datasets cannot provide detailed
informatio n about the s patial distribution of
concentration
in coastal waters [32]. They could only be used to m onitor
on a small- or meso-scale, because of their coarse spatial
resolution
. Instead, some terrestrial observation
satellites with moderate spatial resolutions, such as Landsat,
Système Pour l’Observation de la Terre (SPOT), and Indian
Remote Sensor (IRS), are used for monitoring
concen-
tration in coastal aquatic environment. One of the common
characteristics for these sensors is that they possess high spatial
resolution
but with low signal-to-noise ratio (SNR ,
) and coarse sp ectral resolution [33]. Thus, the
applicability of these sensors in coastal aquatic environment
may be limited by two factors. First, it has long been known
that the bandwidth of absorption trough or scattering peak
of spectral curve related to pigment particle is very narrow
. The coarse spectral resolution makes broadband
sensors insensitive to the absorption or scattering features of
pigment particles, resulting in significant errors; second, the
water-leaving radiance accounts for
or less of the total
radiance measured by the sensor in the blue band and 5% in
the red band. This characteristic requires that the water color
remote sensor must have the high SNR, in order to make the
water-leaving signal detectable theoretically. A lth oug h these
sensors may be considered as the potential sensors for remo te
quantification of
concentration in coastal regions, further
study is still needed to demonstrate the app licability of these
sensors in coastal regions for
concentration estim ation.
It is well known that the coastal waters are the dynamic
aquatic ecosystem s where the bio-chemical event and process
evolve over short temporal scales. In order to detect, monitor,
and predict short term and regional oceanic phenom ena such as
red tides, yellow dust, fishing ground information, sign ificantly
improved SNR performance and spatial and temporal reso-
lutions are essential for ocean color remote sensing satellite.
Fortunately, Geostationary Ocean Color Im ager (GOCI) has
been developed to provide a monitoring of ocean color around
the Korean Peninsula from geostationary platforms. GOCI, the
first ocean color imager to operate from geostationary orbit,
is designed to prov ide multi-spectral data to detect, monitor,
quantify, and predict short-tem changes of coastal ocean envi-
ronment for marine science research and application purpose
[34]. Its mission is to significantly improve ocean observation
from low orbit serv ice by providing high frequency coverage.
Comparing with other ocean color satellites such as MODIS,
GOCI has a unique capability to observe the ocean and coastal
waters with high sp atial resolution (500 m) and temporal resolu-
tion (refresh rate: 1 hour). However, such advantages also meet
some technical challenges. The main challenges are radiom etry
and environment constraints. For ex amp le, geo stationary orbit
is much further from target than low orbit, reducing available
signal more than 500 times [35]. Thus, much more attention
stillshouldbepaidonGOCIinthefuture.
III. A
TMOSPHERIC CORRECTION FOR COASTAL WAT ER S
Generally, the water-leaving reflectance detected by satellite
sensor is fairly weak. For example, the component of the mea-
sured reflectance, backscattering out of the w ater and transmit-
ting to the top of the atmosphere is less than 10% in th e blue
band, and typically much smaller in the green and near-infrared
bands [36]. In order to acquire accurate
concentration
using remote sensing images, the impact of atmospheric scat-
tering and absorption should be accurately removed. According