Satellite detection of harmful algal bloom occurrences in Korean waters

Abstract

Cochlodinium polykrikoides (p) is a planktonic dinoflagellate known to produce red tides responsible for massive fish kills and thereby serious economic loss in Korean coastal waters, particularly during summer and fall seasons. The present study involved analyzing chlorophyll-a (Chl-a) from SeaWiFS ocean color imagery collected over the period 1998–2002 to understand the spatial and temporal aspects of C. polykrikoides blooms that occurred in the enclosed and semi-enclosed bays of the Korean Southeast Sea. NOAA-AVHRR data were used to derive Sea Surface Temperature (SST) to elucidate physical factors affecting the spatial distribution and abundance of C. polykrikoides blooms. The time series of SeaWiFS-derived Chl-a gave an impression that recent red tide events with higher concentrations appeared to span more than 8 weeks during summer and fall seasons and were widespread in most of the Korean Southeast Sea coastal bays and neighboring oceanic waters. Coupled eutrophication and certain oceanic processes were thought to give rise to the formation of massive C. polykrikoides blooms with cell abundances ranging from 1000 to 30,000 cells ml−1, causing heavy mortalities of aquaculture fish and other marine organisms in these areas. Our analysis indicated that Chl-a estimates from SeaWiFS ocean color imagery appeared to be useful in demarcating the locality, spatial extent and distribution of these blooms, but unique identification of C. polykrikoides from non-bloom and sediment dominated waters remains unsuccessful with this data alone. Thus, the classical spectral enhancement and classification techniques such as Forward Principal Component Analysis (FPCA) and Minimum Spectral Distance (MSD) to uniquely identify and better understand C. polykrikoides blooms characteristics from other optical water types were attempted on both low spatial resolution SeaWiFS ocean color imagery and high spatial resolution Landsat-7 ETM+ imagery. Application of these techniques could capture intricate and striking patterns of C. polykrikoides blooms from surrounding non-bloom and sediment dominated waters, providing improved capability of detecting, predicting and monitoring C. polykrikoides bloom in such optically complex waters. The result obtained from MSD classification showed that retrieval of C. polykrikoides bloom from the mixed phase of this bloom with turbid waters was not feasible with the SeaWiFS ocean color imagery, but feasible with Landsat-7 ETM+ imagery that provided more accurate and comparable spatial C. polykrikoides patterns consistent with in situ observations. The dense phase of the bloom estimated from these imageries occupied an area of more than 25 km2 around the coastal bays and the mixed phase extended over several hundreds kilometers towards the Southeast Sea offshore due to exchange of water masses caused by coastal and oceanic processes. Sea surface temperature analyzed from AVHRR infrared data captured the northeastward flow of Tsushima Warm Current (TWC) waters that provided favorable environmental conditions for the rapid growth and subsequent southward initiation of C. polykrikoides blooms in hydrodynamically active regions in the Korean Southeast Sea offshore.

Full text

Page 1

Satellite detection of harmful algal bloom occurrences in
Korean waters
Yu-Hwan Ahna,*, Palanisamy Shanmugama,
Joo-Hyung Ryua, Jong-Chul Jeongb
aSatellite Ocean Research Laboratory, Korea Ocean Research and Development Institute,
Ansan P.O. Box 29, Seoul 425-600, Republic of Korea
bDepartment of Geoinformatics Engineering, Namseoul University, 21 Maeju-ri,
Choongnam 330-800, Republic of Korea
Received 1 February 2005; received in revised form 18 June 2005; accepted 21 July 2005
Abstract
Cochlodinium polykrikoides (p) is a planktonic dinoflagellate known to produce red tides responsible for massive fish kills
and thereby serious economic loss in Korean coastal waters, particularly during summer and fall seasons. The present study
involved analyzing chlorophyll-a (Chl-a) from SeaWiFS ocean color imagery collected over the period 1998–2002 to
understand the spatial and temporal aspects of C. polykrikoides blooms that occurred in the enclosed and semi-enclosed bays
of the Korean Southeast Sea. NOAA-AVHRR data were used to derive Sea Surface Temperature (SST) to elucidate physical
factors affecting the spatial distribution and abundance of C. polykrikoides blooms. The time series of SeaWiFS-derived Chl-a
gave an impression that recent red tide events with higher concentrations appeared to span more than 8 weeks during summer
andfallseasonsandwerewidespreadinmostoftheKoreanSoutheastSeacoastalbaysandneighboringoceanicwaters.Coupled
eutrophication and certain oceanic processes were thought to give rise to the formation of massive C. polykrikoides blooms with
cell abundances ranging from 1000 to 30,000 cells ml?1, causing heavy mortalities of aquaculture fish and other marine
organisms in these areas. Our analysis indicated that Chl-a estimates from SeaWiFS ocean color imagery appeared to be useful
in demarcating the locality, spatial extent and distribution of these blooms, but unique identification of C. polykrikoides from
non-bloom and sediment dominated waters remains unsuccessful with this data alone. Thus, the classical spectral enhancement
andclassificationtechniquessuchasForwardPrincipalComponentAnalysis(FPCA)andMinimumSpectralDistance(MSD)to
uniquely identify and better understand C. polykrikoides blooms characteristics from other optical water types were attempted
on both low spatial resolution SeaWiFS ocean color imagery and high spatial resolution Landsat-7 ETM+ imagery. Application
of these techniques could capture intricate and striking patterns of C. polykrikoides blooms from surrounding non-bloom and
sediment dominated waters, providing improved capability of detecting, predicting and monitoring C. polykrikoides bloom in
such optically complex waters. The result obtained from MSD classification showed that retrieval of C. polykrikoides bloom
from the mixed phase of this bloom with turbid waters was not feasiblewith the SeaWiFS ocean color imagery, butfeasiblewith
www.elsevier.com/locate/hal
Harmful Algae 5 (2006) 213–231
* Corresponding author. Tel.: +82 31 400 6129; fax: +82 31 400 6139.
E-mail addresses: yhahn@kordi.re.kr, pshanmugam@kordi.re.kr (Y.-H. Ahn).
1568-9883/$ – see front matter # 2005 Elsevier B.V. All rights reserved.
doi:10.1016/j.hal.2005.07.007

Page 2

Landsat-7 ETM+ imagery that provided more accurate and comparable spatial C. polykrikoides patterns consistent with in situ
observations. The dense phase of the bloom estimated from these imageries occupied an area of more than 25 km2around the
coastalbaysandthemixedphaseextendedoverseveralhundredskilometerstowardstheSoutheastSeaoffshoreduetoexchange
ofwatermassescaused bycoastal andoceanicprocesses. SeasurfacetemperatureanalyzedfromAVHRRinfrareddatacaptured
thenortheastwardflowofTsushimaWarmCurrent(TWC)watersthatprovidedfavorableenvironmentalconditionsfortherapid
growth and subsequent southward initiation of C. polykrikoides blooms in hydrodynamically active regions in the Korean
Southeast Sea offshore.
# 2005 Elsevier B.V. All rights reserved.
Keywords: C. polykrikoides blooms; Ocean color; SeaWiFS; Landsat-7 ETM+; FPCA; MSD; Korea
1. Introduction
Phytoplankton blooms appear to be an increasingly
common phenomena on a worldwide scale and are
sometimes considered to be harmful either because of
their potential threat to human health through the
consumptionofcontaminatedseafood,asinthecase of
many toxic phytoplankton blooms or through the
changes in species abundance and distributions (often
includingspeciesofcommercialvalue),asinthecaseof
recent brown or red tide blooms (Buskey et al., 1996;
Franks, 1997; Tester and Steidinger, 1997; Cho et al.,
2000; Matsuyama et al., 2001; Sierra-Beltran et al.,
2004;Chengetal.,2005).Thebloomformationoccurs,
whentherateofphytoplanktongrowthexceedstherate
of cell dispersion, due to enhanced anthropogenic
nitrification,whichisoneofthemostpervasivechanges
altering coastal environments worldwide (Shumway,
1990; Burkholder, 1998) or some times steepening of
nutricline by intervenes of physical phenomena
(Falkowski et al., 1991; Olaizola et al., 1993). The
semi-enclosed nature of Korean Southeast Sea coastal
bays often reaches extremity in eutrophic state by
receiving terrestrial wastewater and pollutants related
to heavy rainfall and surface runoff, resulting in
occurrence and outbreaks of series of harmful
Cochlodinium polykrikoides (p) blooms, which appear
to have significantly increased in frequency, intensity
and geographic distribution and caused massive
mortalities of aquaculture fish off the southern and
eastern coasts of Korea (Kim et al., 1999). Among
several dinoflagellate species, C. polykrikoides has
been identified as causative organism of red tides
responsibleformassivefishkillsinwarmtemperateand
tropical waters (Steidinger and Tangen, 1997). Its
monospecific nature with a unique set of conditions
results in high rates of respiration that can cause
dissolved oxygen concentrations to fall to a level low
enough to endanger marine life (Buskey et al., 1996).
Besides coastal eutrophication, other changing envir-
onmental conditions such as warming, resuspension of
spores and advection are also principal factors for the
increased occurrence of monospecific C. polykrikoides
blooms in Korean coastal waters (Kim et al., 1990).
Although there are historical records of this red tide
forming algae since the sixth century, the first
scientific report on red tides was published by Park
and Kim (1967), and later, many red tide events have
been documented along with an extensive red tide
phytoplankton bloom in summer 1995, which caused
heavymortalitiesofaquaculturefishandamountedtoa
loss of ?US$ 95.5 million (Kim et al., 1997). A
similar event of C. polykrikoides in the southern China
Sea during March and April 1998 has been reported to
cause tremendous damage to the coastal aquaculture
industry, wiping out 1500 tonnes of farmed fish, which
was equivalent to half of the entire Hong Kong
aquacultural production of 1997 (Anderson, 1998;
Tang et al., 2003). Perhaps, this species was first
reported from the Caribbean Sea along the southern
coast of Puerto Rico (Margalef, 1961) and northern
Atlantic waters along the American east coast,
Barnegat bay, New Jersey (Silva, 1967). Although
C. polykrikoides is a known red tide species associated
with extensive fish kills and great economic loss in
Korean and Japanese coastal waters (Kim et al., 1999;
Yuki and Yoshimatsu, 1989; Fukuyo et al., 1990), the
mechanism of toxicity and toxin production associated
with extensive aquaculture fish and mollusk damage
have yet to be elucidated (Taylor et al., 1995).
However, secretion of some ichthyologic substances
by C. polykrikoides species has been reported as a
possible cause for fish kills (Onoue et al., 1985;
Hallegraeff, 1992).
Y.-H. Ahn et al./Harmful Algae 5 (2006) 213–231214

Page 3

It is very important to mitigate the impacts of such
harmful algal blooms and therefore there is a need to
monitor the blooms and to forecast their development
and movement (Stumpf, 2001). Because of the high
temporal and spatial heterogeneity of coastal oceanic
ecosystem and processes as well as the expense of
research vessels, monitoring of the development and
movement of these blooms with traditional field
sampling at discrete locations remains critical. Thus,
the only effectiveway of monitoring such blooms on a
regular basis is through remote sensing. Utilization of
remote sensing technology because of its synoptic and
repeat coverage has been explored for detecting
harmful algal blooms, delineating their spatial extent
and addressing their impacts as well as describing the
associated hydrographic conditions (Haddad, 1982;
Cullenetal.,1997;TesterandStumpf,1998;Schofield
et al., 1999). Steidinger and Haddad (1981) were
among the first to demonstrate the potential of satellite
ocean color sensor with Coastal Zone Color Scanner
(CZCS) for the detection of a major Karenia brevis
bloom in the western Florida waters. With the
availability of daily imagery from the current
SeaWiFS operational ocean color sensor, routine
monitoring has been initiated by Stumpf (2001), who
indicates that the use of chlorophyll data might
provide a means for the detection of algal blooms in
coastal waters. Similarly, Chang et al. (2001)
exploited the usefulness of SeaWiFS ocean color
imagery for detecting and monitoring G. catenatum
bloom in New Zealand waters. On the other hand,
satellite-derived sea surface temperature using Adv-
anced Very High Resolution Radiometer (AVHRR)
data, which provides real-time capability with two
thermal infrared channels, has been widely used for
the derivation of circulation patterns, structure of
oceanic fronts, behavior of eddies/meanders and the
location of upwelling zones and associated chemical
and biological features (Pearce and Pattiaratchi, 1997;
Cipollini et al., 1998; Chang et al., 2002; Tang et al.,
2003).
During several research cruises conducted in the
Southeast Sea coastal waters of Korea during August/
September 1998–2002, we observed elevated chlor-
ophyll concentrations as the result of massive C.
polykrikoides blooms triggered by enhanced levels of
nutrients, resulting from heavy rainfall and surface
runoff as well certain oceanic processes. In this study,
we first examine chlorophyll-a (Chl-a) concentrations
analyzed from SeaWiFS ocean color image data of
August/September of the period 1998–2002 in relation
toidentifyingpotentialareasofC.polykrikoidesbloom
occurrences in these waters. We then attempt to
implement certain image enhancement/classification
techniques such as Forward Principal Component
Analysis (FPCA) and Minimum Spectra Distance
(MSD) classification on SeaWiFS image data to
distinguish C. polykrikoides blooms from non-bloom
and sedimentdominated waters.As a part of this study,
apotentialuseofhighresolutionLandsat-ETM+image
datatoidentifythesebloomsisalsoexploredwiththese
techniques. Finally, the retrieved bloom patterns are
corroboratedwiththeinsituobservationsmadeasapart
of this study and by National Fisheries Research and
Development Institute (NFRDI).
2. Materials and methods
2.1. Study area and cruise measurements
The study area includes enclosed and semi-
enclosed bays of the Korean Southeast Sea, encom-
passing complex interactions between physical, che-
mical, biological and geological processes in the
shallow waters with depth less than 60 m, connecting
to the East China Sea (ECS) in the southwest and East
Sea (ES) in the northeast (Fig. 1). The Tsushima Warm
Current (TWC) is known to be a major oceanic current
feature of this region transporting nutrient-abundant
Kuroshio water into the East Sea through the
Tsushima Strait. It splits into two distinct branches, a
northward flow of the East Korean Warm Current
(EKWC) along the east Korean coast and an eastward
flow of the Offshore Branch (OB) along the Japanese
coast (Chang et al., 2002). This warm current appears
to not only provide favorable conditions for the rapid
growth of C. polykrikoides but also controls the spatial
distribution of this bloom in the offshore waters. The
C. polykrikoides blooms since 1981 appear to have
occurred all along these coasts, including Chonsu and
Mokpo in the east and Yoja, Kamak, Chinju, Kosong
and Jin-hae in the south, and Onsan, Ulsan and Yongil
in the east, which seriously undergo eutrophication,
therefore, dinoflagellate C. polykrikoides blooms
spread frequently every year. Fig. 1 shows important
Y.-H. Ahn et al./Harmful Algae 5 (2006) 213–231215

Page 4

locations (color shaded circles) of C. polykrikoides
bloom occurrences with high frequency, intensity and
spatial distribution between summer and fall seasons,
with a peak in August and September. Fig. 2a and b
show red tide waters dominated by C. polykrikoides
bloom in August 1999.
The R/V Olympic cruises were conducted over the
Jin-hae bay and neighboring waters coinciding with
the occurrences of C. polykrikoides blooms during
August/September of the period 1998–2003 (see
Fig. 1). The ship surveys consisted of over 200
stations inside and outside the blooms. Seawater
samplescollected simultaneouslywith the radiometric
measurementswere analyzed for Chl-a concentrations
using Perkin-Elmer Lambda 19 dual-beam spectro-
photometer. The chlorophyll concentrations varied
from 3 to 65 mg m?3around the Jin-hae bay (Fig. 2c
and d), except for a few locations close to the coast
where it reached 207 mg m?3in August 1999 (not
shown here). Fig. 2e shows absorption spectra of
phytoplankton aph(l) measured in the red tide waters
around the Jin-hae bay, which exhibit two dominant
absorption peaks around 444 and 670 nm and the
magnitude of these peaks increases with the increase
in Chl-a concentration. On the other hand, remote
sensing reflectance Rrs(l) spectra corresponding to
chlorophyllconcentrations
absorption maxima at 445 nm and reflectance maxima
at 567 nm and sun-induced chlorophyll fluorescence
peak maxima at 688 nm (Fig. 2f). When Chl-a
concentration increases, the magnitude of the fluor-
escence peak also increases with a notable decrease
towards the blue part of the spectrum. Note that the
position of the peak remains constant, offering a way
to estimate chlorophyll-a concentrations and detect
these blooms using the relationship between this
signal and Chl-a concentrations (Ahn and Shanmu-
gam, submitted for publication).
18–34 mg m?3
show
Y.-H. Ahn et al./Harmful Algae 5 (2006) 213–231216
Fig. 1. Study area map shows potential areas of harmful C. polykrikoides bloom occurrences (color shaded circles) along the Korean southern
and eastern coasts. The arrow mark indicates areas of in situ measurements carried out over the Jin-hae bay and surrounding waters during
August/September of the period 1998–2003. In addition, regional surface current patterns observed from the satellite-tracked drifter trajectories
in the Korean and neighboringwaters (Lie et al., 1998) are illustrated showing TsushimaWarm Current (TWC) and its branches the East Korean
Warm Current and Offshore Branch.

Page 5

Y.-H. Ahn et al./Harmful Algae 5 (2006) 213–231217
Fig.2. (aandb)RedtidewatersdominatedbyC.polykrikoidesspeciesintheJin-haebayduringAugust1998.C.polykrikoidesisfoundinsingle
cells (ellipsoidal shape) and chain form. (c and d) SurfaceChl-a concentration measured in August1998 and 1999.(e) Absorption spectra aph(l)
determined in the red tide waters of the Jin-hae bay exhibiting two absorption features around 444 and 670 nm. (f) Example of remote sensing
reflectance Rrs(l) spectra, corresponding to chlorophyll concentrations 18–34 mg m?3, showing absorption maxima at 445 nm and reflectance
maxima at 567 nm and chlorophyll fluorescence maxima around 688 nm.

Page 6

2.2. Satellite imagery and processing
Level 1A SeaWiFS products were collected from
our KORDI satellite data receiving station during
August/September of the period 1998–2002. These
data were coincident with the rapid growth and
outbreaks of C. polykrikoides blooms in the Jin-hae
bay and neighboring waters. The ocean color
radiances were atmospherically corrected and pro-
cessed to level 2 using NASA SeaDAS version 4.4
(Tomlinson et al., 2004), which has an updated
atmospheric correction algorithm that compensates
for near-infrared water-leaving radiance and for
absorbing aerosols (Stumpf et al., 2003). The surface
Chl-a concentrations were then derived by using
NASA OC4v4 bio-optical algorithm within SeaDAS
software (Yoder et al., 2002). The SeaWiFS-derived
Chl-a was nearly consistent with the in situ Chl-a data
collected outside the coastal bays. Similarly, NOAA-
AVHRR imageries were geo-referenced to a common
grid and projection system at a spatial resolution of
1.1 km and the land pixels were subject to masking to
a single value. Sea surface temperature was then
computed by combining the radiance temperatures
derived from the two individual thermal bands to
account for the varying amounts of water vapor in the
atmosphere (Barton, 1995). This was accomplished
with the split window Multi Channel Sea Surface
Temperature (MCSST) dual channel algorithm (avail-
able with TerraScan software) because it was found to
be more efficient for the accurate SST computation,
with the root mean square differences of about 0.6 8C
(Li et al., 2001).
In order to map accurately the initiation and
movements of C. polykrikoides blooms, SeaWiFS and
Landsat-7 ETM+ image data were processed sepa-
rately with the classical spectral enhancement and
classification techniques such as Forward Principal
Component Analysis and Minimum Spectral Distance
(MSD) classifier. These techniques are, in principle,
the most widely used techniques for extracting land
coverinformationonthesurfacefromremotelysensed
data (Smara et al., 1998; Oetter et al., 2000),
aquaculture form information from Landsat-5 TM
data in the Korean coastal waters (Shanmugam et al.,
2004) and also detecting algal blooms from SeaWiFS
image data in the North Sea (Pasterkamp et al., 2002).
The essence of FPCA is that it uses a linear
transformation of multi-band data to translate and
rotate data set from the actual measurement space,
such as radiance in spectral bands, to a new
measurement space that maximizes the variance of
the original data set. The method considers all the
available measured initial differences at once and
provides, in prioritized order, sets of coefficients to
rotate the input bands to new dimensions of maximum
variation. In contrast, the MSD classifier is a well-
known mathematical decision rule used for hard
classification of remotely sensed image data. It
operates by calculating the spectral distance between
the measurement vector for the pixel to be classified
and the mean vector for each signature (Jensen, 1995;
Schowengerdt, 1997).
Before application of these methods to SeaWiFS
and Landsat-7 ETM+ image data, two steps were
followed: (1) both SeaWiFS and Landsat-7 ETM+
image data were georeferenced to a standard datum
and projection and georeferencing accuracy was less
than 0.5 pixel, and (2) pixel digital counts were
converted to radiance at the top of the atmosphere
(TOA) using SeaWiFS calibration coefficients (Hoo-
ker et al., 1994) and Landsat-7 ETM+ calibration
coefficients (http://ltpwww.gsfc.nasa.gov/IAS/hand-
book/handbook_toc.html), respectively. It was fol-
lowed by atmospheric correction of these images with
the Spectral Shape Matching Method (SSMM)
developed by Ahn and Shanmugam (2004).
In order to produce a single map of likeliest class,
the MSD required a number of data points (also
referred to as training samples) for each class.
Therefore, a training data set was constructed from
the image consisting of a group of prototypical data
points that represent the data characteristics of each
class, spanning the observed variability in feature
values while maintaining class separability in feature
space. These data points from SeaWiFS image
consisted of all eight bands and Landsat-7 ETM+
image of four bands (three visible and one near
infrared red bands). In order to evaluate the signature
separability between the selected training data points,
two statistical measures of distance were reported,
namely Transformed Divergence (TD) and Jeffries–
Matusita (JM) (Swain and Davis, 1978). These values
have lower and upper bounds, i.e., 0 and 2000 for TD
and 0 and 1414 for JM distance. Here we considered
these values ranging from 0 and 2. If the calculated
Y.-H. Ahn et al./Harmful Algae 5 (2006) 213–231218

Page 7

distance is equal to the upper bound then the
signatures can be said to be totally separable in the
bands being analyzed. A calculated distance of zero
means that the signatures are inseparable (Swain and
Davis, 1978).
2.3. Harmful algal bloom (red tide) events in
Korean coastal waters
Since 1970, the geographical distribution of red
tide algal blooms was limited to the enclosed bays
such as the Jin-hae and Kosong bays, but they turned
out to be widespread to Kangnung waters in the East
Sea, Inchon and Mokpo bordering the Yellow Sea and
Yoja bay in the western part of the South Sea (Kim
etal., 1990).Aseries ofred tidealgalbloom outbreaks
was recorded in Korean coastal waters as a part of red
tide monitoring program implemented by National
FisheriesResearchand
(NFRDI) under the support of Ministry of Maritime
Affairs and Fisheries (MOMAF) (Fig. 3). The annual
total number of red tide outbreaks reveals that 8 events
occurred in 1981, encompassing 3 harmful diono-
flagellate C. polykrikoides blooms with cell concen-
trations 1500 cells ml?1, and 21 in 1982, and followed
by 28 harmful events (with cell concentrations over
27,000 cells ml?1) out of 65 in 1995, the year of C.
polykrikoides blooms. Recently, the C. polykrikoides
blooms have affected most of coastal areas, with some
cases causedby more than one toxic algal species such
as Prorocentrum micans, Gymnodinium mikimotoi,
Gymnodinium sanguineum, Prorocentrum minimum
and Heterosigma akashiwo. The toxins of these
species have been reported to cause extensive
DevelopmentInstitute
mortality in fish and invertebrate populations (Buskey
et al., 1996; Kim et al., 1994). Table 1 characterizes
red tide blooms into four stages based on species
toxicity, bloom density, area and duration. It gives an
impression that recent red tide events with higher
density appear tohave persisted formore than 8weeks
and are widespread in most of the enclosed bays and
neighboring ocean waters.
3. Results
3.1. Distribution of chlorophyll during August/
September
The variability of SeaWiFS-derived Chl-a and
AVHRR-derived SST throughout the Jin-hae bay and
neighboring waters is displayed in Fig. 4a–f, where
elevated Chl-a concentrations are observed with
striking patterns at several places around coastal
and offshore waters of the Southeast Sea during
August/September of the period 1998–2000. With the
exception of river mouths and estuaries where artifact
occurs due to presence of other water constituents, the
observed patterns in other areas may be caused by C.
polykrikoides bloom, which tends to spread east-
wardly and dominate euphotic waters off the Ulsan
coast in August 1998 (see attached figure on the top
corner of Fig. 4a). SeaWiFS-derived Chl-a concentra-
tionduring thisperiod appears tohavevariedfrom3to
54 mg m?3inside the Jin-hae bay and 0.6–14 mg m?3
outside the bay. The accumulation of C. polykrikoides
within the euphotic layer off the Ulsan coast isthought
to result mainly from vertical mixing and enhanced
Y.-H. Ahn et al./Harmful Algae 5 (2006) 213–231219
Fig. 3. Number of observations of harmful algal bloom outbreaks off the Korean coast since 1980. Note that the frequency of the flagellates
bloom outbreaks appears to have increased over time to that of diatoms bloom outbreaks (Kim, 1998).

Page 8

levels of essential nutrients pumped upward from the
bottom layer, which subsequently leads to the
formation of quick eutrophic situations off the Ulsan
coast. A time series of SeaWiFS-derived Chl-a
suggests that the bloom event lasted several weeks
(2 months) inside the Jin-hae bay while spanning 1–2
weeks (a short-term variability) in the nutrient-rich
upwelled waters. The upwelled waters closely linked
with EKWC (Byun, 1989) are characteristic with
temperature typical of 2–10 8C surrounded by warm
current waters of 7–23 8C and renowned for their high
fisheries production in these areas. Blooms of high
intensity were previously reported in such coastal
upwelling areas at the mid and high latitudes (Glover
and Brewer, 1988; Morel and Berthon, 1989). Though
nutrient-rich upwelled waters were characterized by
Y.-H. Ahn et al./Harmful Algae 5 (2006) 213–231220
Table 1
Characterization of red tide algal blooms in Korean coastal waters since 1980
Terms Before 19801981–19881989–1992 1993–1997
Species toxicity
Affected area
Bloom density (cells/ml)
The longest duration
Harmless
Partial area
1000
1 week
Harmless/harmful
Partial/wide area
1000–5000
1–2 weeks
Harmful
Widespread/South Sea
2000–10000
3 weeks (81)
Harmful
Widespread overall coast
3000–30000
8 weeks (1995)
Fig. 4. (a–f) Spatial and temporal aspects of surface chlorophyll concentrations analyzed from the SeaWiFS ocean color imagery showing
recurrent patterns of C. polykrikoides blooms (a–d) caused by the intervenes of several physical processes, which are manifest in the sea surface
temperature images from AVHRR infrared data (e and f). (g) In situ data show the spatial distribution of C. polykrikoides bloom in the
southeastern coastal areas during August 1998–2000.

Page 9

the surface salinity and potential temperature obtained
from CTD sensors, there was no data available for
corroborating the accumulation of C. polykrikoides
species off the Ulsan coast, because most of our in situ
data were confined to the Jin-hae bay and its
surrounding waters during August. However, Sea-
WiFS-derived Chl-a was consistent with in situ data
between Kosong and western coast of Geoje island
(see Fig. 4g (top)).
In contrast, the SeaWiFS-derived Chl-a on 25
September 1999 demonstrates the southward exten-
sionofC.polykrikoidesblooms,favoredbytheflowof
TWC waters along the Tsushima Strait in the South
Sea offshore (Fig. 4a and e). The Chl-a elevation
during the southward penetration of this bloom is due
to the fact that the growth rate of C. polykrikoides is
maximum 0.41 day?1at 25 8C and salinity 34 (Kim
et al., 2004), who support the hypothesis that C.
polykrikoides species blooms extensively with a peak
when water temperature increases from 22 to 25 8C
due to northeastward intrusion of TWC along the
Tsushima Strait. This warm current not only provides
favorableenvironmentalconditionsforthisbloom,but
also derives oceanic nutrients from the Kuroshio,
conveying for the seed of the offshore bloom. The
NFRDI data indicate that the cell abundance in the
coastalareasbetweenGeojeislandandKosongranged
from 500 to 26,000 cells ml?1, these levels are enough
to produce stress and weakness in aquaculture fish and
increased susceptibility to other infections. For the
bloom, SeaWiFS-derived Chl-a concentration ranged
from 3 to 56 mg m?3inside the Jin-hae bay and 2–
23 mg m?3outside the bay, closely resembling in situ
Chl-aatseveralstationsexceptsomelocationscloseto
the coast where it reached very high levels (compar-
ison made for August 1999 images because of having
in situ match-up). According to the CTD survey
results, the southward extension of C. polykrikoides
bloom was associated with low saline and cold water
(u < 14 and S < 34.3) around the Jin-hae bay, with
pronounced hydrographic variability at both surface
and subsurface which was indicative of the thermoha-
line water flow along the Tsushima Strait in the South
Sea offshore. It is conspicuous in AVHRR-derived
SST image (Fig. 4e), which elucidates temperature
variability between TWC and surrounding waters. A
close inspection on SeaWiFS-derived Chl-a on 27
September 1999 reveals that, in a matter of 2 days, the
southward extension pattern of C. polykrikoides
bloom transformed eastwardly by the intensification
of the northward intrusion of TWC (see Fig. 4e where
TWC appears to flow along the northern periphery of
the island in the South Sea offshore), resulting in a
mushroom-like structure in the SeaWiFS-derived Chl-
a with 1.5–17 mg m?3(Fig. 4b). The early stage of
this bloom captured from SeaWiFS-derived Chl-a
compared well with in situ data from Suh et al. (2004)
(Fig. 4g (middle)).
On 27 September 1999, the appearance of C.
polykrikoides bloom can also be evident around
Pohang coastal waters (Fig. 4b). This bloom spanning
few weeks was apparently triggered by anthropogenic
nutrients driven by the Yongil river, while a massive
phytoplankton bloom observed in waters north off
Pohang coast appears to be associated with the eddy-
like feature indicated by a circle in Fig. 4b and e. This
eddy was typically episodic and energetic enough to
cause an injection of essential nutrients into the
euphotic layer, resulting in the enhanced levels of
algal biomass which produced Chl-a concentrations
from 0.9 to 2.7 mg m?3. Falkowski et al. (1991)
reported such increased biomass concentration asso-
ciated with the cyclonic eddy features. The eddy-like
feature, that did not trap algal matter-like rings,
produced upwelling over a time scale sufficiently long
to produce a transient bloom at this location and
resulted in a trail-like Chl-a pattern that extended over
a few hundreds kilometers from waters north off
Pohang coast towaters north off Ulleungdo in the East
Sea. Trajectory of a satellite-tracked surface drifter
(Agrosbuoy) andhydrographic
demonstrated that this pattern was induced as a
consequence of EKWC along the east Korean coast
that produced anticyclonic circulation of water mass
around Ulleungdo with high speed of about 50–
60 cm s?1(Chang et al., 2002).
In September 2000, C. polykrikoides bloomed
extensively with higher cell concentrations and
discolored waters all along the southeastern coast,
posing potential threat to aquaculture fisheries and
other marine organisms (Fig. 4c and d). The coastal
blooms initiated in cold and low saline waters
resulting from monsoonal rains, while the rapid
growth and offshore extension of C. polykrikoides
bloom was sustained by the physical warm currents.
The TWC characterized by S > 32.2 and u > 21 8C at
measurements
Y.-H. Ahn et al./Harmful Algae 5 (2006) 213–231221

Page 10

surface and S > 34.4 and u > 15 8C at 75 dbar was
very intensive during this period directing the spatial
patterns of these blooms (over hundreds of kilometer)
on the Offshore Branch along the Japanese coast
(Fig. 4c, d and f). In fact, this was the first time such
massive blooms directed on the Offshore Branch were
detected over an expanse stretching hundreds of
kilometeralongtheJapanesecoast.Theabsence ofsea
truth data did not allow confirming this bloom event.
CTD observations showed that the southward pene-
tration of NKCC waters along the Korean coast
induced hampering the escalation of EKWC, giving
rise to the strong eastward flow of the Offshore
Branch, which directed the spatial structure of this
bloom to the southern East Sea during September
2000 (Fig. 4d and f). As a result of NKCC waters, a
previously formed anticyclonic eddy feature in May
2000 appeared to be dominated by waters of lower
salinity and temperature S < 33 and u < 15 8C, and
consequentlySeaWiFS-derivedChl-aaroundtheeddy
feature south off Ulleungdo became very low 0.5–
2.1 mg m?3(as seen in Fig. 4d), indicative of the
deformation of this eddy feature in the East Sea during
September 2000.
3.2. Detection of red tide blooms with the spectral
enhancement and classification techniques
The SeaWiFS-derived Chl-a combined with phy-
sical oceanographic information provided a means of
detecting, delineating and monitoring of coastal and
offshore extent of the C. polykrikoides blooms in the
Korean Southeast Sea. However, analysis of Chl-a
alone does not seem to be effective for the delineation
of C. polykrikoides blooms, because it helps only in
identifying areas with high chlorophyll concentration
rather than distinguishing C. polykrikoides blooms
from surrounding non-bloom and sediment dominated
turbid waters. Tang et al. (2003) observed that such
high-chlorophyll concentrations often result from
artifacts owing to abundance of dissolved organic
and particulate inorganic sediments around river
mouths and estuaries. Therefore, it is very essential
to correctly identify C. polykrikoides blooms from
other optically dominant water types, which requires
additional capability rather than simply identifying
chlorophyll patterns (Stumpf et al., 2003). Optical
methods have been previously developed for detecting
Coccolithophores and Trichodesmium spp. (Brown
and Yoder, 1994; Subramaniam and Carpenter, 1994),
but these methods require water lacking particulate
matter and other dissolved pigments.
Withtheintension ofcorrectly identifyingpotential
areas of C. polykrikoides blooms and enhanced
understanding of bloom patterns,
Principal Component Analysis and Minimum Spec-
tral Distance (MSD) classification techniques were
attempted on both low spatial resolution SeaWiFS
ocean color imagery (acquired on 14 August 1998 and
19 September 2000) and high spatial resolution Land-
sat-7 ETM+ imagery (acquired on 24 August 2001).
During the process of FPCA, a covariance matrix of
the input bands was computed, and this covariance
matrix was used in the subsequent stage to compute
the principal components or eigenvectors. Table 2
shows the eigenvalues and variance (%) explained by
each component computed from SeaWiFS images
acquired on 14 August 1998 and 19 September 2000,
respectively. The total variance in percentage was
calculated for each component using the following
formula: % variance explained = (eigenvalues of
component ? 100)/sum
%variancei¼ li? 100=Pn
critical data contained in the eight bands. Similarly for
components 1–8, the % variance explained decreased
rapidly and the cumulative percentage that is a check
for all eight components was equal 100%. The last
three components containing mostly noise were
disregarded here. For the convenience, multiple
components (containing least to most information)
the Forward
ofall eigenvalues,i.e.,
i¼1li. It appears that the
first five components contain nearly 100% of the
Y.-H. Ahn et al./Harmful Algae 5 (2006) 213–231222
Table 2
The eigenvalues and % variance explained by each component
computed from SeaWiFS images acquired on 14 August 1998
and 19 September 2000
ComponentSeaWiFS image
(14 August 1998)
SeaWiFS image
(19 September 2000)
Eigenvalue% Variance Eigenvalue% Variance
1
2
3
4
5
6
7
8
56.46536
1.1767
0.118244
0.022436
0.019221
0.006051
0.000833
0.000301
97.67547
2.035491
0.204543
0.03881
0.03325
0.010468
0.001441
0.000521
54.68769
0.493074
0.172289
0.028757
0.016871
0.004643
0.000233
0.000103
98.70771
0.889967
0.31097
0.051905
0.030452
0.00838
0.00042
0.000186

Page 11

were displayed simultaneously to facilitate interpreta-
tion of C. polykrikoides bloom from other dominant
optical water types.
Fig. 5a and b compare SeaWiFS-derived Chl-a and
spectrally transformed FPCA image obtained from
multiplied components 1 and 5 (containing most and
least information). Comparison reveals that SeaWiFS-
derived Chl-a simply shows chlorophyll pattern that
was thought to result from C. polykrikoides blooms,
but FPCA image supports improved detection and
discrimination of C. polykrikoides blooms (in pink
tone) from non-bloom (in blue to green) and turbid
waters (brown). Because of optical complexity of
these waters and problem associated with the standard
SeaWiFS atmospheric correction algorithm, a white
mask was created on the SeaWiFS-derived Chl-a in
the Jin-hae bay and neighboring coastal bays (Fig. 5a).
In contrast, the FPCA image processed after applica-
tion of Spectral Shape Matching Method (SSMM)
(Ahn and Shanmugam, 2004) for correcting the
atmospheric effects in the SeaWiFS image exhibit
apparent patterns of C. polykrikoides bloom in these
coastal bays(Fig. 5b).This meansthat SSMM enabled
retrieval of water-leaving radiance in high scattering
and absorbing waters in proximity to the coast. From
the FPCA image, it is evident that C. polykrikoides
bloom around the Jin-hae bay and a small pocket of
turbid water around Pusan river mouth appear to move
eastward to influence optical properties of water
masses off the Ulsan coast. The large accumulation of
C. polykrikoides in the upwelling area off the Ulsan
coast might have caused damage to the aquaculture
farms and associated marine ecosystem. However, our
data set restricted to the Jin-hae bay did not confirm
this bloom event in the Ulsan coastal waters and
offshore waters.
Similarly, Fig. 5c and d show the SeaWiFS-derived
Chl-a and a color composite FPCA image from three
components 5, 2 and 1 that demonstrated majority of
spatial variability among C. polykrikoides bloom and
Y.-H. Ahn et al./Harmful Algae 5 (2006) 213–231223
Fig.5. (a–f)DetectionofC. polykrikoides bloomswiththespectralenhancementtechniquesappliedto SeaWiFS andLandsat-7ETM+data. (a–
d)ComparisonbetweenSeaWiFS-derivedChl-aandFPCAimagesand(e andf)originalandenhancedimages(contraststretching)ofLandsat-7
ETM+ data acquired over the Southeast Sea on 24 August 2001.

Page 12

other water constituents. It appears that C. poly-
krikoides bloom patterns inferred from the SeaWiFS-
derived Chl-a compare well with those (orange and
red patches) displayed in the FPCA image, excluding
areas of turbid plume associated with the Pusan river
mouth and its adjacent coastal waters and coastal
waters west off Kosong coast, where SeaWiFS-
derived Chl-a falsely identifies C. polykrikoides
blooms. Interpretation of Chl-a suggests that C.
polykrikoides bloom indications in areas of such
highly colored waters should generally be ignored
(Stumpf et al., 2003). In FPCA image, five categories
can be distinguished, namely dense red tides (in
yellow and orange), less dense red tides (in pink and
red), highly turbid water (in blue), less turbid water (in
dark blue) and non-bloom waters (in green) (Fig. 5d).
The movement and future location of the dense phase
of C. polykrikoides bloom can also be captured from
an enhanced color composite zoomed image (using
original image of B841) shown on the right side of
Fig. 5e (denoted by a cross mark). The dense phase
moved southwardly occupies large areas of Kosong
waters, posing potential threat to the entire Kosong
coastal ecosystem during 19 September 2000. On the
contrary, the C. polykrikoides bloom appeared to be
highly persistent around the Jin-hae bay and expanded
significantly to encompass areas in the Southeast Sea
offshore and southern East Sea. The highly dynamic
C. polykrikoides bloom patterns outside the bay
coexisted with turbid waters from the Pusan river, and
the intrusion of Tsushima Warm Current remained the
main cause governing the geographical distribution of
the C. polykrikoides blooms in these waters.
The high resolution Landsat-7 ETM+ imagery
(only visible and near infrared bands) acquired on 24
August 2001 over the Southeast Sea was also
processed to provide what information can be gained
from such high resolution data regarding C. poly-
krikoides bloom dynamics, its direction of movement
and areas of high persistence and distribution. This
was the period of highly intense C. polykrikoides
blooms coinciding with our field measurement and
field and aerial survey (by helicopter) conducted by
NFRDI. The field measurement revealed that C.
polykrikoides blooms appeared to be darker with
several striking patterns observed around the Kosong
waters, Jin-hae bay and off the southern and eastern
coasts of Geoje island. These dark patches of C.
polykrikoides bloom were difficult to be detected with
the geometrically and atmospherically corrected False
Color Composite (FCC) image (generated using
ETM+ bands 432) (Fig. 5e), but were conspicuous
when the spectral enhancement was done by
manipulating the range of digital radiance values in
the color composite images (Fig. 5f is displayed with
B321 and a small subset shown on the top right corner
of this image is displayed with B123), graphically
represented by their histograms. The multiple color
display consisted of considerable spatial variability
characterized by three different color ranges corre-
sponding to different levels of red tide algal species
and associated water constituents. The subset image
shown on the top right corner of Fig. 5f demonstrates
linear and filament-like dark patches (originating from
the coastal areas of Kosong) consistent with NFRDI
observations (see Fig. 7c). These dark patches were
also tracked from SeaWiFS-derived Chl-a (4–53 and
1–7 mg m?3for less dense phase) image of this
period. The development of most part of this bloom
mighthavebeencausedbycoupledeutrophicationand
oceanic processes.
In order to map potential areas of C. polykrikoides
blooms, the use of Minimum Spectral Distance
classification was explored using atmospherically
corrected SeaWiFS and Landsat-7 ETM+ imagery.
Pasterkamp et al. (2002) established algal bloom
classes with the application of supervised classifica-
tion techniques using SeaWiFS imagery, while
Danaher and Omongain (1992) developed a similar
classification algorithm based on the singular value
decomposition (SVD) technique for the detection and
classification of algal bloom types from reflectance
data. The success of these classifications largely
depends on the knowledge of the study area and
optical significance of various water types. In this
study, several training data points (samples) were
identified for each of four dominant water class types
in case of SeaWiFS image (Fig. 6a–d) and three
dominant water class types in case of Landsat-7
ETM+ image in order to perform MSD classification
of both these image data. Attention was paid in
selecting these training data points so as to reduce
possible errors associated with classification. The
mean spectral values (water-leaving radiance in units
of mW cm?2mm?1s?1) of these classes in each of
SeaWiFS and Landsat-7 ETM+ bands are given in
Y.-H. Ahn et al./Harmful Algae 5 (2006) 213–231224

Page 13

Table 3. Prior to the classification, these training
signatures were evaluated using two separability
measures—Jeffries–Matusita
Divergence (TD) and the results are shown in
Table 4. It is confirmed that the selected training data
(JM), Transformed
points are good separable from each other having high
values in both JM and TD measures.
The result of classification of SeaWiFS image
indicates that good discrimination was possible
between the dense phase of C. polykrikoides bloom
Y.-H. Ahn et al./Harmful Algae 5 (2006) 213–231225
Fig. 6. (a–d) Training data points (samples) collected from the atmospherically corrected SeaWiFS image data (19 September 2000) in order to
performMSDclassification,showingmajoropticalwatertypessuchasredtide,mixedtype(redtideandsuspendedsediments),turbidwaterand
clear water. The mean spectral values in each SeaWiFS bands and for each class are given in Table 3. X-axis—SeaWiFS band number and Y-
axis—water-leaving radiance.
Table 3
MeanspectralvaluesineachSeaWiFSandLandsat-7ETM+bandsfordifferenttrainingdatapointsrepresentingmajoropticalwatertypesinthe
study area
Band no.SeaWiFS image (19 September 2000) (mean spectrum)Landsat-7 ETM+ image (24 August 2001)
(mean spectrum)
Red tide Mixed typeTurbid water Clear waterRed tideTurbid water Clear water
1
2
3
4
5
6
7
8
0.3344
0.3517
0.3495
0.3787
0.6055
0.2163
0.0764
0.0524
0.6080
0.9176
1.2711
1.3715
1.6205
0.5548
0.0786
0.0463
1.1987
1.8321
2.5105
2.4713
2.3824
0.7244
0.2147
0.1531
0.8803
0.9350
0.7302
0.5726
0.3558
0.1411
0.0696
0.0509
0.2468
0.3711
0.2039
0.0686
0.7138
1.1691
1.0439
0.1229
0.7151
0.4780
0.1268
0.0394

Page 14

and non-bloom and sediment dominated waters, but
mixed phase of this bloom coupled with turbid plume
(propagating from Pusan coastal area to its offshore)
could not be fully captured from non-bloom waters.
The possible reason is that the MSD decision rule
calculates the spectral distance between the measure-
ment vector for the pixel to be classified and the mean
vector for each signature, therefore, a class with less
variance, like blue waters, may tend to become
overestimated, since the pixels that belong to the class
are usually spectrally closer to their mean (Jensen,
1995). The inadequate spatial resolution of the
SeaWiFS image results in the large errors with the
mixed water type class (Fig. 7a). In the classified
image, red color indicates C. polykrikoides bloom,
thistle color represents highly turbid waters, green
color stands for mixed type and blue color denotes
clear water. The predominantly high-suspended
sediments resulted from the river and process of
sediment resuspension due to tidal currents and
bottom circulations. The estimated area for the dense
phase of the C. polykrikoides bloom was about 26 km2
covering the Jin-hae bay and Kosong waters, while it
was over 30 km2for the mixed phase of the bloom
extending to include offshore waters in the Southeast
Sea.
In contrast, classification of Landsat-7 ETM+
image data with MSD mapping technique could bring
out detailed information about the location, direction
of movement, distribution and intricate patterns of C.
polykrikoides bloom off the Koreansoutheastern coast
(Fig. 7b). Three distinct categories as red tide,
suspended sediments and seawater were mapped
from Landsat-7 ETM+ image, because intension with
the derivation of more number of classes produced
unrealisticresultsandledtotheunderestimation ofthe
spatial extent of C. polykrikoides bloom in these
waters. Comparison between classification of low and
high spatial resolution data demonstrated that classi-
fication accuracy was higher in Landsat-7 ETM+
image than in SeaWiFS image data, however, both
images clearly showed strong persistence of C.
polykrikoides bloom between Geoje island and
Kosong coastal areas and inside the Jin-hae bay.
For validation of the classification, we used data
collected on 21 August 2001 through field and aerial
survey (by helicopter) by NFRDI (Fig. 7c) (Suh et al.,
2004). The C. polykrikoides bloom density was
observed to exceed 3000 cells ml?1(in red color) in
the western and southern parts off Geoje island, while
relatively less dense bloom with concentrations less
than 3000 cells ml?1(in stripped polygons) domi-
nated Kosong coastal waters and all along the
southeastern coastal areas (Fig. 7c). It should be
recalled that bloom of higher density (exceeding
1000 cells ml?1) would have a potential impact on the
aquaculture fish and other marine organisms (see
Table 1) (Kim, 1998). Validation of classification
revealed that the inferred patterns and distribution of
C. polykrikoides bloom by MSD classification on
Landsat-7 ETM+ image data closely resembled to that
observed in situ and through aerial reconnaissance
survey by NFRDI, indicating the reliability of the
MSD classification.
A similar classification can also be achieved
through direct spectral discrimination using feature
Y.-H. Ahn et al./Harmful Algae 5 (2006) 213–231 226
Table 4
Results of spectral separability analysis performed on different
trainingdatapointsfromSeaWiFSandLandsat-7ETM+imagedataa
SeaWiFS radiance image
(19 September 2000)
Landsat-7 ETM+ image
(24 August 2001)
Red tide
Mixed: (2.0, 2.0)
Turbid water: (2.0, 2.0)
Clear water: (2.0, 2.0)
Turbid water: (2.0, 2.0)
Clear water: (1.9, 1.9)
Mixed
Red tide: (2.0, 2.0)
Turbid water: (2.0, 2.0)
Clear water: (2.0, 2.0)
Turbid water
Red tide: (2.0, 2.0)
Mixed: (2.0, 2.0)
Clear water: (2.0, 2.0)
Red tide: (2.0, 2.0)
Clear water: (2.0, 2.0)
Clear water
Red tide: (2.0, 2.0)
Mixed: (2.0, 2.0)
Turbid water: (2.0, 2.0)
Red tide: (1.9, 1.9)
Turbid water (2.0, 2.0)
Pair separation (least to most)
Red tide and mixed: (2.0)
Mixed and turbid water: (2.0)
Red tide and clear water: (2.0)
Red tide and turbid water: (2.0)
Mixed and clear water: (2.0)
Turbid and clear water: (2.0)
Red tide and turbid
water: (2.0)
Red tide and clear
water: (1.9)
Turbid water and clear
water: (2.0)
aSeparability measures: Jeffries–Matusita (JM), Transformed
Divergence (TD).

Page 15

space shown in Fig. 8. The ratio Lw(510/555) aga-
inst Lw(443) (from SeaWiFS) allows one to extract
patterns of red tides from other dominant optical
water types such as turbid waters and non-bloom
(clear waters) waters. However, the success of
this method largely depends on verification data to
confirm the bloom event, because non-red tide
bloom may also show similar patterns in the feature
space.
4. Discussion and conclusion
The increased terrestrial and water-born pollu-
tants in the closed and semi-enclosed bays of the
Korean Southeast Sea result in the occurrence of
potentially toxic red tide blooms, seasonal anoxia and
shellfish intoxication (Kim, 1998). C. polykrikoides
which is a species with a very adoptable physiology,
frequently occurring, widespread and persistent and
Y.-H. Ahn et al./Harmful Algae 5 (2006) 213–231227
Fig. 7. (a and b) Detection of red tide algal blooms with the MSD mapping technique applied to SeaWiFS (19 September 2000) and Landsat-7
ETM+ images (24 August 2001). These images give an idea that C. polykrikoides bloom recurs in the Jin-hae bay and Kosong coastal areas west
ofGeoje islandduringthe fall season.We observed thatthe highspatialresolutionLandsat-7ETM+imagebringsoutdetailedinformation about
the C. polykrikoides bloom patterns consistent with in situ data (c) collected on 21 August 2001 by NFRDI (Suh et al., 2004). No data are
available for corroborating massive blooms occurred in the Jin-hae bay during this period.