ArticlePDF Available

Hydrodynamic Tracking of the Massive Spring 1998 Red Tide in Hong Kong


Abstract and Figures

In subtropical coastal waters around Hong Kong, algal blooms and red tides have been frequently observed over the past two decades. In particular, in March-April 1998, a massive red tide invaded the northeastern and southern coastal waters of Hong Kong. The devastating red tide resulted in the worst fish kill in Hong Kong's history, the most significant impacts being at the Lo Tik Wan and Sok Kwu Wan fish culture zones on Lamma Island. This work reports the first scientific investigation of the cause of this massive red tide. A calibrated three-dimensional (3D) hydrodynamic model for the Pearl River Estuary, Delft3D, is applied to study the advective transport of red tides. Based on the tidal boundary conditions and the measured wind data for a typical spring season, the 3D flow field is computed and extensive surface drogue tracking performed for releases in different parts of the coastal waters and for different tidal and wind conditions. The results show that a bloom initiated in Mirs Bay (Nan Au or Tap Mun) in the northeastern water would likely be transported to the southern coastal waters under the combined action of tidal current and wind. The computed bloom tracking patterns are generally supported by observations and are consistent with the temporal and spatial patterns of individual fish kill events in the 1998 red tide. We conclude that the major cause of the bloom being transported into the southern waters and East Lamma Channel (and causing the massive fish kill) is the generally strong wind in March-April 1998 and the change in wind direction in early April under almost diurnal tidal conditions. Further, it is most probable that the red tide originated in Mirs Bay rather than from outside Hong Kong. The findings provide a firm basis for environmental and fisheries management.
Content may be subject to copyright.
Hydrodynamic Tracking of the Massive Spring 1998
Red Tide in Hong Kong
Joseph H. W. Lee1and Bo Qu2
Abstract: In subtropical coastal waters around Hong Kong, algal blooms and red tides have been frequently observed over the past two
decades. In particular, in March–April 1998, a massive red tide invaded the northeastern and southern coastal waters of Hong Kong. The
devastating red tide resulted in the worst fish kill in Hong Kong’s history, the most significant impacts being at the Lo Tik Wan and Sok
Kwu Wan fish culture zones on Lamma Island. This work reports the first scientific investigation of the cause of this massive red tide. A
calibrated three-dimensional 3Dhydrodynamic model for the Pearl River Estuary, Delft3D, is applied to study the advective transport of
red tides. Based on the tidal boundary conditions and the measured wind data for a typical spring season, the 3D flow field is computed
and extensive surface drogue tracking performed for releases in different parts of the coastal waters and for different tidal and wind
conditions. The results show that a bloom initiated in Mirs Bay Nan Au orTap Munin the northeastern water would likely be transported
to the southern coastal waters under the combined action of tidal current and wind. The computed bloom tracking patterns are generally
supported by observations and are consistent with the temporal and spatial patterns of individual fish kill events in the 1998 red tide. We
conclude that the major cause of the bloom being transported into the southern waters and East Lamma Channel and causing the massive
fish killis the generally strong wind in March–April 1998 and the change in wind direction in early April under almost diurnal tidal
conditions. Further, it is most probable that the red tide originated in Mirs Bay rather than from outside Hong Kong. The findings provide
a firm basis for environmental and fisheries management.
DOI: 10.1061/ASCE0733-93722004130:5535
CE Database subject headings: Hong Kong; Coastal management; Fisheries; Disasters; Hydrodynamics; Hydraulic models; Water
quality; Eutrophication; Tracking.
Hong Kong is situated at the mouth of the Pearl River Estuary in
Southern China Fig. 1. In subtropical coastal waters around
Hong Kong and South China, algal blooms and red tides due to
the rapid growth of microscopic phytoplanktonare often ob-
served. Under the right environmental conditions favorable tem-
perature, solar radiation, nutrient concentration, predation pres-
sure, wind speed, and tidal flushingthese blooms can occur and
subside over rather short time scales—on the order of days to a
few weeks. Algal blooms can lead to discoloration of the marine
water, which may lead to beach closures, severe dissolved oxygen
depletion, fish kills, and shellfish poisoning. Over the past two
decades, massive fish kills due to oxygen depletion have been
observed in some of the marine fish culture zones in Hong Kong
Lee et al. 1991; a few toxic algal blooms have also been re-
ported. In particular, in April 1998, a devastating red tide due to
the dinoflagellate Karenia digitatumresulted in the worst fish kill
in Hong Kong’s history—it destroyed over 80% 3,400 tof cul-
tured fish stock, with estimated loss of more than HK$312 million
Dickman 1998; Yang et al. 2000. It appears that the dinoflagel-
late species produced copious amounts of a sticky mucus which
coated the fish gills, leading to suffocation and mortality within
hours after the fish farms were hit by the red tide Dickman
1998. Other explanations are also possible, such as the produc-
tion of fish toxins Anderson 1998.
The 1998 red tide was truly an unusual event compared with
all past blooms in Hong Kong waters. It was the most serious red
tide in Hong Kong’s history. It is recorded that 30 red tide inci-
dents appeared in the first six months of 1998 with the most
serious events during the period 19 March to 17 April Anderson
1998. Whereas the general ecological response of phytoplankton
to environmental conditions has been extensively studied, the
causality and dynamics of algal blooms are extremely compli-
cated and not well understood. In particular, for the massive red
tide of 1998, the press reports and water quality measurements
suggest a migration of a red tide initiated in mid-March in the
northeastern waters of Hong Kong, in Mirs Bay where red tides
in Hong Kong are usually first sighted in any given year, which
drifted to the southern coastal waters with tidal currents Fig.
2a兲兴. There is also speculation that the red tide may have initiated
from outside Hong Kong, in Daya Bay along the South China
coast. Alternatively, the observations may also reflect merely a
change of environmental conditions in the various bloom loca-
tions during March–April 1998, with in situ growth occurring at
different times along the coast. Asystematic study of the probable
cause of this massive red tide is clearly desirable; a general un-
derstanding of the hydrodynamic transport of algal blooms would
be most useful for environmental and fisheries management.
1Professor, Dept. of Civil Engineering, Univ. of Hong Kong,
Pokfulam Rd., Hong Kong, China.
2Postdoctoral Research Fellow, Dept. of Civil Engineering, Univ. of
Hong Kong, Pokfulam Rd., Hong Kong, China.
Note. Associate Editor: A. Bruce DeVantier. Discussion open until
October 1, 2004. Separate discussions must be submitted for individual
papers. To extend the closing date by one month, a written request must
be filed with the ASCE Managing Editor. The manuscript for this paper
was submitted for review and possible publication on February 20, 2002;
approved on June 5, 2003. This paper is part of the Journal of Environ-
mental Engineering, Vol. 130, No. 5, May 1, 2004. ©ASCE, ISSN 0733-
This paper reports a detailed study of the role of tidal hydro-
dynamics in the 1998 red tide. Particular emphasis is placed on
the intense red tide observed in southern waters and associated
fish kills in the fish farms located off the East Lamma Channel—
namely, the Lo Tik Wan and Sok Kwu Wan fish culture zones,
which suffered unprecedented complete losses. First, the three-
dimensional 3Dhydrodynamic model Delft3D is fully tested
against analytical solutions of long wave propagation in rectan-
gular and radial channels of varying bathymetry. The numerical
model previously calibrated with extensive field data for the
Pearl River Estuaryis then employed to study the tidal circula-
tion during the average dry and wet seasons Hong Kong has a
Fig. 1. Pearl River Estuary model for algal bloom tracking
Fig. 2. aReported red tide movement during March 18April 14 1998 from newspaper.bClose-up view of Hong Kong Island and Lamma
distinct dry season with little rain from November to March, and
a wet season from June to September frequented by typhoons and
monsoons; the other months are transition periods. Extensive nu-
merical drogue tracking experiments are then carried out using a
refined grid for the northeastern waters of Hong Kong. The
drogue tracks are thoroughly studied for releases in different lo-
cations in the northeastern waters, and for the full range of pos-
sible scenarios release times, tidal range, wind conditions. The
results for both the typical March–April months and the particu-
lar conditions prevailing in spring 1998 are compared. Represen-
tative tracks are presented to illustrate the general nature of hy-
drodynamic transport of algal blooms in Hong Kong. The most
probable cause of the spring 1998 bloom is discussed in relation
to the hydrodynamic circulation and drogue transport patterns.
Hydrodynamic Modeling of Hong Kong’s Coastal
Governing Equations
The three-dimensional hydrodynamic model Delft3D is employed
for this study. Delft3D solves the Reynolds-averaged turbulent
flow equations for an incompressible fluid using a hydrostatic
pressure assumption. The resulting horizontal momentum and
continuity equations for a Cartesian coordinate systemfor baro-
clinic flow are as follows the shallow water equations:
⫽⫺ 1
⳵␴ fu
⫽⫺ 1
where (u,v)horizontal velocities in the (x,y) directions, re-
spectively; free surface elevation above mean sea level;
dmean depth; fCoriolis parameter; Ppressure force term,
and Srefers to the source/sink term to simulate discharges and
withdrawal. The depth-averaged velocities are
In the vertical direction, the coordinate dimensionless depthis
with Htotal water depth. At the bottom ⫽⫺1, and at the sur-
face 0. The coordinate system is boundary fitted in the ver-
tical plane. In the coordinate system, the horizontal length scale
is assumed to be much larger than the water depth.
The vertical velocity in the coordinate system is computed
from the continuity equation
⳵␴ ⫽⫺
The vertical velocity win the x-y-zCartesian coordinate system
can be expressed in terms of the horizontal velocities, water
depths, water levels, and vertical coordinate according to
The hydrostatic pressure is given by
where density of water computed as a function of salinity Sin
pptand temperature Tin °Cvia an equation of state. The hori-
zontal pressure gradients are obtained by using Leibnitz’s rule
where 01,023 kg/m3reference water density applicable to
Hong Kong waters. The forces Fxand Fyin the momentum equa-
tions 1and 2represent the horizontal Reynold’s stresses
where Hhorizontal eddy viscosity coefficient which is much
larger than the vertical eddy viscosity. The horizontal viscosity
coefficient His defined by H2D3Dmol and the vertical
eddy diffusivity coefficient v3Dmol . Here the 2D part 2D
is associated with the contribution of horizontal motions and forc-
ings that cannot be resolved sub-grid-scale turbulenceby the
horizontal grid; this coefficient is obtained by calibration. The 3D
part 3D is referred to as the 3D turbulence and is computed
following the standard two-equation turbulence model see
below.mol is the molecular viscosity.
Transport of Conservative Constituents
The transport of salt and heat is modeled by a conservative trans-
port equation
where Dhhorizontal diffusivity; and Scsource/sink term.
Turbulence Closure
A standard two-equation turbulence model is used for turbulence
closure Launder and Spalding 1974; Rodi 1980. The transport
equation for the turbulence kinetic energy kand the rate of dissi-
pation of k() are nonlinearly coupled via the production and
dissipation terms and the eddy diffusivity Dv. The transport
equations for kand are given by
The production terms Pk,Pand buoyancy source terms Bk,B
are defined as
⳵␴ (12c)
where the empirical model calibration constants are adopted as
c11.44, c21.92, and c31.0 for stable stratification and
c30.0 for unstable stratification.
The vertical eddy viscosity vVreferred to earlier as 3D)is
determined by
with the empirical constant c0.09. The horizontal eddy diffu-
sivity DHand vertical eddy diffusivity DVare assumed the same
as the corresponding viscosities.
Boundary Conditions
In practical applications such as the Pearl River Estuary model
Fig. 1, the model uses orthogonal curvilinear coordinates 共␰,␩兲.
Application of orthogonal curvilinear coordinates yields a bound-
ary fitting in the horizontal plane. The free slip formulation is
adopted. The boundary conditions for the free surface, the seabed,
and the open boundary are as follows.
Vertical Boundary Condition
In the coordinate system, the free surface and bottom corre-
spond to a plane. The vertical boundary conditions for the ver-
tical velocity are given by
10 and 00 (14)
Bed Boundary Condition
At the seabed, the boundary conditions for the momentum equa-
tion are
where bbottom shear stress; and bx and byxand ycompo-
nents computed from the velocities in the first layer above the bed
via a standard quadratic law.
The turbulence quantities are prescribed via a local equilib-
rium assumption, with
where u*b
b/;the von Ka
´n constant0.41; and
z0roughness height obtained by calibration. The logarithmic ve-
locity profile is assumed in the bottom boundary layer
The parameter z0can be linked to the actual geometric roughness
as a fraction of the RMS value of the subgrid bottom fluctuations.
For rough walls, z0ks/30, where ksNikuradse roughness
length scale and is determined experimentally.
Surface Boundary Condition
At the free surface, the boundary conditions for the momentum
equations are
cos (17a)
sin (17b)
where angle between the wind stress vector and the local di-
rection of the grid line yconst. The magnitude of the wind shear
stress is determined by a quadratic expression
where adensity of air; Cdwind drag coefficient; and
U10wind speed 10 m above the free surface. In the present
study, the measured wind speed recorded by the Hong Kong Ob-
servatory at Waglan Island, at Z55.8 m above mean sea level
MSL, is adopted. This wind speed is corrected to U10 using an
empirical relation that has proved reliable by analysis of historical
data Chin and Leong 1978
U/U100.2330.656 log10Z4.75(18b)
where Zelevation above MSL. The wind drag coefficient Cdis
assumed to vary linearly with wind speed from 0.00063 to
0.00723 in the range of U100 –100 m/s Delft Hydraulics 1997.
In the presence of wind, the turbulent kinetic energy and its
dissipation rate are specified as
where zsroughness length obtained by calibration.
Open Boundary
Fig. 1 shows the Pearl Estuary model grid. Along open bound-
aries, water levels are specified by a time history of water levels
using a number of tidal constituents derived from long-term tidal
data east, south, and west boundaries. A linear vertical concen-
tration profile is specified at the open boundaries for an inflow
Numerical Solution
A staggered grid is used for the discretization of the horizontal as
well as the vertical gradients; the water level pressureand con-
centration salinitypoints are defined in the center of a cell,
whereas the velocity components are defined on the cell faces.
The horizontal diffusion terms are centrally differenced while an
upwind scheme is used for the advection terms. For the time
integration of the horizontal diffusion term, the Crank-Nicholson
method is applied; source terms are integrated explicitly. The so-
lution of the horizontal momentum and continuity equation pro-
ceeds via a cyclic implicit procedure using the alternating direc-
tion implicit scheme Stelling 1984. In the vertical direction, the
fluxes are discretized with a central scheme; the time integration
is fully implicit. Further numerical details can be found in Stelling
1984and Delft Hydraulics 1997.
The Delft3d model used in this study has been thoroughly
tested against analytic solutions of long wave propagation Ap-
pendix I. For cases in which the analytical model with linearized
bottom friction is applicable Lynch and Gray 1978, the com-
puted elevations and velocities are in very good agreement with
the analytical solutions.
Numerical Model Results of Pearl Estuary Model
Pearl Estuary Grid
Fig. 1 shows the Pearl River Estuary model set up for studying
the tidal circulation and water quality of Hong Kong’s coastal
waters. A 206141 boundary-fitted orthogonal grid is used, with
ten uniformly distributed vertical layers; the grid in the northeast-
ern waters is specially refined to study algal bloom transport pat-
terns. Tidal forcing is applied along the eastern, western, and
southern boundaries; nine tidal constituents derived from an ex-
tended harmonic analysis of long term tidal dataare used to
generate a time history of predicted water levels at the open
boundaries; they are O1, P1, K1, N2, M2, S2, K2, M4, and MS4.
The mean sea level conditions are set at the different boundary
stations from a low of 1.18 m in the southwest of the boundary
station to 1.31 m in the northeast of the boundary station for the
dry season. The measured salinities for both the dry and wet
seasons are used in the open boundaries. For the dry season, a
vertically uniform salinity of 34 ppt is specified at the ocean
boundaries east, south, westfor inflow, except in the northwest
boundary station surface layer near the Pearl River Estuary, where
S30 ppt is prescribed. Seasonal mean freshwater flows from the
nine outlets of the Pearl River are prescribed at the inflow bound-
aries the locations of five major discharge outlets are indicated in
Fig. 1 and the detailed discharges are listed in Table 1.
The horizontal eddy viscosity and diffusivity 2D component
are all set to 1 m2/s. A Manning coefficient of n0.020 is used
for shallow water; in deeper water greater than 25 m,n
0.026 is used. These values for viscosity/diffusivity and bottom
friction were obtained by calibration and validation against exten-
sive tidal level, velocity, and salinity measurements Delft Hy-
draulics 1997, 1998. The model is run from an initial state ob-
tained by running the model from a cold start; typically the
transients are dissipated within 1–2 tidal cycles.
Tidal Circulation in Hong Kong Waters
The hydrography of Hong Kong’s coastal waters is mainly influ-
enced by three factors: tidal currents, monsoon-affected ocean
currents, and Pearl River discharges. In the dry season, the salin-
ity is approximately vertically homogenous and ranges from
about 34 ppt at the open boundary to 15–20 ppt at the mouth of
the Pearl Estuary near Shenzhen Fig. 3a兲兴. There is no signifi-
cant difference between surface and bottom salinity in winter
Fig. 3. In the wet season, however, tidal currents and the Pearl
River discharge create significant vertical and horizontal salinity
gradients. Compared to the dry season, there is much more sig-
nificant salinity stratification in the western part of the Pearl Es-
tuary Fig. 4. Surface and bottom salinity differentials of up to
7–10 ppt are commonly observed in the northwest areas of Hong
Kong. The eastern waters are relatively sheltered from the Pearl
River; these waters are more oceanic with typical salinity of
32–34 ppt.
In general, tides in Hong Kong are mixed and mainly semidi-
urnal. The range of spring tide and neap tide is from 1.0 to 2.1 m.
Tidal currents are determined by the interaction of ocean tides
with the local bathymetry. In the dry season, the currents in Hong
Kong waters are mainly from SE to NW through Victoria Harbor
and northwest up the Pearl River during flood Fig. 5a兲兴 and
from Pearl River Estuary to southeast water during ebb Fig.
5b兲兴. The tidal flow in the East Lamma Channel Figs. 5a and
b, close-up viewis mainly from the southeast to northwest di-
rection during flood and in the opposite direction during ebb.
Tidal currents south of Lamma Island generally flow from the
northeast to the southwest direction. Fig. 6 shows the detailed
flow field at midflood and beginning of ebb in Mirs Bay. It can be
seen that the flow in the northern part of Mirs Bay is mainly
characterized by an anticlockwise circulation. In the south of the
bay, the circulation is more likely in a clockwise direction. Both
types of circulation patterns are present in flood and ebb.
In the wet season, the flow pattern not shownis heavily
influenced by the Pearl River and is much more complicated than
the dry season flow, with significant vertical stratification.
Salinity-driven circulation is evident. The flow is mainly from
west to east. The residual currents are mainly from NW to SE in
Victoria Harbor and western areas, although currents in both di-
rections can also be found through northwestern waters and east
of the Lamma Channel. In southern waters, the flows are mainly
from W/SW to E/NE.
This study mainly focuses on the effect of the tidal circulation
and wind on the 1998 red tide that occurred toward the end of the
dry season.
Algal Bloom Tracking
Red Tide Tracking Model
The aim is to study the transport of an algal bloom initiated in the
northeast Hong Kong waters as frequently observed. In particu-
lar, the relation between tidal circulation and the reported se-
quence of events from March 18 to mid-April 1998 Fig. 2aand
Table 2is studied thoroughly. Initial extensive computations
using the original coarserPearl Estuary grid led to the adoption
of a much more refined grid for the northeastern waters including
Table 1. Dry Season Pearl River Discharges from Nine River Outlets
River outlet Discharge
Humen 795
Jiaomen 700
Hongqili 225
Hengmen 441
Muodaomen 1,145
Jitimen 205
Hutiaomen 240
Aimen 365
Deep Bay 2.5
Mirs Bay, Tolo Harbor, and Port Sheltershown in Fig. 1; the
computations are performed using two, four, eight, and ten layers.
In the area of interest northeast and southern waters, the grid
size varies from 70 m to 1.75 km; in the ocean waters next to the
western open boundary, a coarse grid up to 6 km is used. A time
step of t1 or 2 min is adopted, with a Courant number
gh/x) in the range of 0.241.8. The tidal circulation is
computed for the typical dry and wet seasons and extensive
drogue tracking experiments are performed. The massive red tide
of 1998 is then examined using the actual wind data from Febru-
ary to April 1998. The daily-averaged wind data at Waglan Island
Hong Kong Observatory 1998, 1999are used.
Fig. 3. Computed salinity in dry season: aSurface and bbottom salinity
Fig. 4. Computed salinity in wet season: aSurface and bbottom salinity
The drogue tracking simply follows the path of a floating ob-
ject transported by the surface current i.e., the current in the
upper layer. In the drogue tracking computations, the drogue is
released after 5 days flow simulation, when the flow has reached
a dynamic steady state for the typical dry and wet seasons.In
view of the extensive drogue tracking simulations that need to be
made on the entire Pearl Estuary grid, the number of vertical
layers has been suitably reduced. Computations have been made
with a two-layer model 0.5:0.5 vertical distributionand a four-
layer model with 0.1:0.2:0.3:0.4 vertical distribution, and the
results compared with those from an eight-layer model Appendix
II. Based on the results a two-layer model is considered inad-
equate; while a four-layer model gives accurate results and retains
reasonable computational economy for the large amount of
drogue tracking.
An intensive study in drogue tracking for March–April 1998 is
undertaken using a four-layer Pearl Estuary grid, with a Courant
number of around 0.8. As a drogue track of interest would depend
on the release location as well as the exact time of release, it
would be futile to enumerate the drogue tracks released in all
locations within northeastern waters during the months of Febru-
ary, March, andApril of 1998—or the general dry season months.
We thus proceed heuristically. Generally, a drogue is released
from the surface layer at the middle of a grid—for release posi-
tions throughout Mirs Bay, Tolo Harbor, and Daya Bay, see Fig.
7. Numerical experiments show that the drogue path for such a
release can almost describe the general path of movement of a
bloom initiated anywhere within the same grid, or within the
same region. In general, numerous particles released from differ-
ent positions within one grid would have similar drogue track
paths within Mirs Bay and even down to the southeast of Hong
Kong Island see later discussion, Fig. 11; these drogue tracks
start to diverge only after they arrive at the southeast of Hong
Kong Island. This is generally true in regions where the velocity
Fig. 5. Computed surface velocity in dry season during: aFlood tide and bebb tide
has a definite direction or weak velocity gradients; however, in
regions of significant velocity gradients, particles released from
different positions within the same grid may drift along different
paths. Through such drogue tracking experiments for a large num-
ber of releases at different tidal stages, a good idea of the general
trend of the red tide movement can be gained. The possibility that
a red tide could be transported to the southern waters and East
Lamma Channel can also be reliably assessed.
For a four-layer calculation on a 450 MHz PC, it generally
takes 30 h to run a one-month flow simulation and drogue track-
Massive Red Tide of Spring 1998
According to newspaper reports based on scattered observations
of the fishermen, the 1998 red tide appeared first in the northeast
Mirs Bay. It was sighted in the Kat O area on March 18, 1998 by
the Hong Kong Agriculture and Fisheries Department; the red tide
then appeared to be transported south toward and into Port Shel-
ter, and further westward to Lamma Island, where the most severe
fish kill occurred Fig. 2a兲兴. There is also speculation that the red
tide may have originated from Daya Bay and been transported to
the southeast of Mirs Bay and then northward to the Kat O area.
These reports are not based on actual data or measurements. The
generally overcast sky conditions prevailing in March and April
in Hong Kong also make it difficult to extract the movement of
this massive red tide from satellite photos using remote sensing
techniques; so far, no successful attempt has been reported.
In addition to the informal reports and sightings, cell counts of
dominant algal bloom species were also measured on water
samples collected in fish farms or on beaches by the Agriculture,
Fisheries, and Conservation Department of the Hong Kong Gov-
ernment. Table 2 summarizes all the measurements of the harmful
algal bloom HABspecies Karenia digitatum recorded during
the 1998 red tide. It can be seen that the measurements are gen-
erally consistent with the news reports. The HAB was first mea-
sured in Kat O on March 18; it was subsequently detected at the
Leung Shuen Wan fish culture zone FCZat the mouth of Port
Shelter on March 23, apparently being transported to inner Port
Fig. 6. Computed surface velocity in Mirs Bay for typical dry season
during: aFlood and bebb
Table 2. Measured Algal Cell Counts of the Harmful Algal Bloom Species Karenia digitatum during the Spring 1998 Hong Kong Red Tide
Time Location Cells/mL
March 18, 1998 O Pui Tong Kat O兲共FK3,480
March 23, 1998 Leung Shuen Wan southeast of Port Shelter兲共FK820
April 3, 1998 Kai Lung Wan inner Port Shelter兲共FK2,220
April 8, 1998 Lo Tik Wan East Lamma Channel1,760
April 9, 1998 Tung Lung Chau FK73,700
April 9, 1998 Ma Wan 14,380
April 10, 1998 Cheung Sha Wan Lantau Island兲共FK82,300
April 10, 1998 Tap Mun 4,440
April 11, 1998 Lo Tik Wan FK225,500
April 11, 1998 Yung Shue Au Tolo Harbour5,200
April 12, 1998 Lo Tik Wan East Lamma Channel兲共FK12,620
April 12, 1998 Repulse Bay South of Hong Kong Island34,300
April 12, 1998 Yim Tin Tsai Tolo Habour38,200
April 13, 1998 Chung Hom Kok South of Hong Kong Island17,950
April 13, 1998 Middle Bay South of Hong Kong Island19,075
April 13, 1998 Repulse Bay 28,900
April 13, 1998 Shek O Southeast of Hong Kong Island35,850
April 13, 1998 South Bay South of Hong Kong Island10,560
April 14, 1998 Cheung Chau Tung Wan7,875
April 14, 1998 Chueng Chau Kwun Yam Wan11,510
April 14, 1998 South Bay 29,150
April 14, 1998 Shum Wan Tolo Harbour19,900
April 15, 1998 Kwun Yam Wan; Shek O; Repulse Bay 690; 370;260
April 15, 1998 Lo Tik Wan; Sok Kwu Wan 1; 480;520
Note: Data from Hong Kong Agriculture, Fisheries and Conservation Department. FKfish kill.
Shelter Kai Lung Wan FCZby April 3. By April 89, high
concentrations of the HAB species were detected at Tung Lung,
Ma Wan, and Lo Tik Wan East Lamma Channel. Concentrations
as high as 225,500/mL were recorded on April 11 at Lo Tik Wan
FCZ where all the fish stock was destroyed within a couple of
days. Around April 1213, high concentrations of the order of
10,000/mL were also measured at several important beaches fre-
quented by tourists on the south of Hong Kong Island Repulse
Bay, Chung Hom Kok, Middle Bay, Shek O. The red tide also
appeared to be transported to Cheung Chau west of Lamma Is-
landby April 14. The data suggest that the HAB also appeared in
Tolo Harbor Yung Shue Au on April 11, Yim Tin Tsai on April
12, Shum Wan on April 14at around the same time as or after the
devastating red tide hit Lo Tik Wan during April 11–12.
During this period, red tide outbreaks of other nonharmful
species also occurred, e.g., Porocentrum minimum in Yung Shue
Au and Yim Tin Tsai during March 25–31 and also in early April;
however, these red tides were not associated with fish mortalities
and have not been included in Table 2.
In order to unravel what really happened in spring 1998,
drogue tracking is performed for a release from different locations
in Mirs Bay and at different stages of the tidal cycle and different
wind conditions. This tracking is performed both for the typical
dry season and for spring 1998.
Drogue Tracking for the Dry Season in General
First, the algal bloom tracking is performed using the synthetic
tidal boundary conditions for March and April 1998 but with a
mean wind for the dry season. In general, for the dry season the
average measured wind direction in Hong Kong is 45° northeast
directionand wind speed is U105 m/s Delft Hydraulics 1997,
1998; these typical constantvalues are used. Adrogue/bloom is
released from different locations within Mirs Bay, Tolo Harbor,
and Daya Bay; extensive drogue tracking experiments are per-
formed for a release at different tidal conditions: March 11
spring tide, March 16, March 20 neap tide, and March 27
spring tide, and early April. Based on a detailed study, the fol-
lowing points can be made and are illustrated in Fig. 7, which
shows the drogue tracks for the representative releases. It is found
that a drogue released in the northern part even Kat Oof Mirs
Bay or in Tolo Harbor in the dry season would not likely travel
down south. Typically, the drogue follows a counterclockwise
Figs. 7a–c, h, and m兲兴 or clockwise path Fig. 7g兲兴, and ends
up in the Kat O area, or is further transported from Kat O down to
Tolo Harbor Figs. 7b, c, f, and k兲兴.A bloom in the Kat O region
may originate from Yian Tian Fig. 7a兲兴, Xiao Mei Sha Fig.
7b兲兴, or northeast of Mirs Bay Figs. 7c and g兲兴 69 days earlier.
It is also possible for a bloom initiated in Nan Au, northeast Mirs
Bay, or the Tap Mun area to be transported directly into Tolo
Harbor Figs. 7d, j, and l兲兴. A bloom located slightly east of Kat
OFig. 7i兲兴 or along the north shore near Xiao Mei Sha Fig.
7f兲兴 may be transported toward Starling Inlet. By the nature of
the counterclockwise circulation in Mirs Bay Fig. 6, a drogue
released around the Tap Mun area may move in two directions: it
can move northward toward Tolo Channel if released slightly
north of Tap Mun or southward down in the Tai Long Wan direc-
tion if released slightly south of Tap Mun Fig. 7l兲兴. A bloom
from northeast Mirs Bay would also follow the counterclockwise
circulation toward Xiao Mei Sha, Yian Tian, and the northwest
coast of Mirs Bay Figs. 7m, o, and p兲兴.
In general, it is extremely unlikely that a bloom initiated in the
northern part of Mirs Bay, the Kat O area, or Tolo Harbor would
be transported down to the East Lamma Channel in the typical
dry season. Fig. 7 shows clearly that a bloom from the Nan Au or
Tap Mun area can be transported along a NE/SWpath to Tai Long
Wan Fig. 7j兲兴, or southeast of Lamma Island Figs. 7n–p兲兴.
Typically, blooms transported southward along this path would
drift toward Victoria Harbor and less likely drift to East Lamma
Channel. The time required from Mirs Bay to East Lamma Chan-
nel is listed in Table 3.
Fig. 7. Surface drogue tracking for a release in different parts of Mirs Bay and at different tidal stages in typical dry season arrow indicates
drogue release point; output interval3 days
On the other hand, a bloom from Daya Bay would always be
transported in the southeast direction toward the southeast of
Lamma Island Figs. 7e and q兲兴. It takes 612 days for a bloom
from Daya Bay reach southeast of Lamma Channel. The presence
of a strong wind can reduce the travel time of a surface drogue
Drogue Tracking in Spring 1998
Fig. 8 shows the measured wind speed and direction at Waglan
Island during March and April 1998; the wind speed corrected to
10 m above MSL, U10 , is also shown. It can be seen that the wind
is quite strong at the beginning of March and on March 22–23
and also relatively strong at the beginning of April. The wind
direction is mostly NE during March, but changed direction on
April 1, 4, and 67 to northwest/northerly; the wind direction
then changed in a clockwise manner to the SE direction during
April 7–12. The tidal elevation and velocity in the East Lamma
Channel from mid-March to mid-April are shown in Fig. 9; the
almost diurnal tide during April 3–6 can be noted.
For the actual wind conditions in March–April 1998, similar
extensive numerical drogue tracking as previously described is
performed for releases from different tidal periods in both March
and April. Figs. 10A and Bshow a summary of representative
tracks. As in the dry season, a drogue released fromYian Tian and
Xiao Mei Sha follows a counterclockwise path Figs. 10A—a
and b兲兴 or from Northeast Mirs Bay along a clockwise path Fig.
10A—c兲兴 toward the Kat O area. This suggests that the observed
red tide at Kat O on March 18 Table 2may well have initiated
in Yian Tian, Xiao Mei Sha, or Northeast Mirs Bay around March
10. A bloom initiated northeast of Mirs Bay or Nan Au can also
be transported into Tolo Harbor Figs. 10A—d and f兲兴. The
bloom could drift from the Kat O area to Tap Mun, or pass Tai
Long Wan to southwest of Port Shelter, or move further down
even to Victoria Harbor Fig. 10A—h兲兴. A red tide initiated in
Northeast Mirs Bay, Nan Au, or Tap Mun could have been trans-
ported to Tai Long Wan Figs. 10A—e–g兲兴. This is consistent
with a reported red tide of unknown species in Tai Long Wan on
March 17 not in Table 2. In particular, it appears the red tide in
Leung Shuen Wan on March 23 Table 2may be caused by
blooms initiating from northeast of Mirs Bay around March 16
Fig. 10A—g兲兴. Compared to the typical dry season, the stronger
wind conditions in March–April 1998 increase the chance of
bloom transport toward Lamma Channel. Fig. 10A—ishows
that a bloom could have been initiated at Nan Au on March 16
and been transported to Sok Kwu Wan in East Lamma Channel
on March 24 due to the strong wind during March 22–23 and the
almost diurnal tide period. However, it is very unlikely that red
tide would be transported from north or northeast of Mirs Bay or
Tolo Harbor down to east Lamma Channel in March 1998.
The drogue tracks show clearly that blooms initiated in Nan
Au and Tap Mun in late March or early April can be transported
to East Lamma Island/southern waters several days later. The
bloom could possibly travel along the path from Nan Au to Tap
Mun and then down to the south beaches of Hong Kong Island or
East Lamma Channel Figs. 10A—j, l, m, and o, Figs. 10B—a
and b兲兴. This suggests that the red tide observed in the southern
beaches of Hong Kong Island Repulse Bay, Chung Hom Kok,
Middle Bay, Shek O; see Fig. 2band Table 2on April 1214
mostly likely originated from the Nan Au area Figs. 10B—b, g,
and h兲兴 or the Tap Mun area Fig. 10B—f兲兴 5–10 days earlier. It
is also possible for a bloom from the Nan Au area to pass Tung
Lung into Victoria Harbor and be transported to the Ma Wan area
on April 9 Fig. 10A—pand Table 2. Most importantly, the red
tide at Lo Tik Wan on April 9–12 most likely originated from the
Nan Au area Figs. 10B—a and c兲兴. We note that a bloom could
drift down even from the Kat O area, passing through south of
Port Shelter to southeast of Lamma Island in this special period
Fig. 10B—e兲兴.
Fig. 8. Daily prevailing wind at Waglan Island in March and April
1998 Fig. 9. Tidal elevation and current at East Lamma Channel during
Table 3. Travel Time from Key Locations in Mirs Bay in Typical Dry Season
From Yian Tian KatO Nan Au Nan Au Ko Lau Wan or Tap MunNan Au
To Kat O Tolo Channel Tolo Channel Ko Lau Wan East of Lamma Channel East of Lamma Channel
Days 5–6 3–9 3–12 2–3 3–8 5–10
It is also interesting to see the difference between drogue
tracks in March/early April and those in mid-April 1998. From
March to early April, the blooms would likely move from the Nan
Au area westward toward Tap Mun and then southward toward
East Lamma Channel Figs. 10A—i, j, m, and o; Figs. 10B—a
and b兲兴. When the wind direction changes from NE to SE during
April 9–12, a bloom can travel directly from Nan Au or northeast
of Mirs Bay toward East Lamma Channel without passing the Tap
Mun area Figs. 10B—g–i兲兴. The path of algal bloom during
April 9–12 is more clearly shown in Fig. 13 below gray path.
As in the dry season, it is very unlikely for a bloom initiated in
Daya Bay not shownto enter the East Lamma Channel drogue
moves in southwest direction toward southeast of Lamma Island
within 3–12 days.
Fig. 11 illustrates the difference in possible algal bloom paths
between 1998 and the dry season in general for a group of
Fig. 10. Hydrodynamic transport for a bloom initiated in different parts of Mirs Bay in spring 1998 A and B. Output interval3 days for
A—a-h; 1 day for A—i-pand B—a-j.
drogues released in Mirs Bay. It is clear that blooms initiated in
northeast Mirs Bay or the Tap Mun area have a much greater
tendency to be transported to East Lamma Channel and southern
waters. In spring 1998 blooms from northeast Mirs Bay can be
transported south and then westward toward East Lamma Chan-
nel Fig. 11a1兲兴, while for the typical dry season these blooms
would more likely be transported to the north part of Mirs Bay or
Tolo Harbor. Based on all of the above results we conclude that
there are three possible sources for the observed HAB in East
Lamma Channel in spring 1998: 1Nan Au; 2Tap Mun area;
and 3northeast of Mirs Bay.
Role of Hydrodynamic Transport
Based on a study of numerous drogue tracks and hypothesis test-
ing, the unusually strong wind in March 1998 and early April and
the change in wind direction from NE to SE during early April
Fig. 8appear to play a major role in the massive red tide ob-
served in the southern waters of Hong Kong in April 1998. In
particular, the change in wind direction in early April resulted in
the creation of a clockwise gyre in Mirs Bay—which in turn
greatly facilitated the transport of blooms toward Lamma Island
and southern beaches. The change of wind direction from NE to
SE during April 9–12 also enhances the transport of the red tide
into the East Lamma Island around April 11.
Fig. 12 shows the typical surface velocity field for the north-
eastern waters and approach to Hong Kong and Lamma Islands
for both ebb and flood tides during April 7–12, 1998. It is seen
that, contrary to the characteristic counterclockwise circulation in
Mirs Bay Fig. 6, a clockwise gyre can be found in Mirs Bay.
This results in a distinct high velocity belt extending diagonally
from the Nan Au area to Port Shelter and southeast of Lamma
Island Figs. 12a2 and b2and Fig. 13. Within this belt, the
velocity is much larger than in the surroundings; the southward
velocity is also considerably stronger than that for the dry season
in general. In Figs. 12c1 and c2the computed surface current in
Fig. 11. Comparison of paths of groups of drogues for 1998 and
typical dry season
Fig. 12. Typical surface velocity field for April 1998 during period of southeast winds April 7–12;rightcomparison with typical dry season
the middle of Mirs Bay Location Aand the mouth of Port Shel-
ter Location Bin spring 1998 dashed lineis compared with
that for the dry season in general solid line. It can be seen that
during spring 1998 the southward negativevelocity is signifi-
cantly higher than in the dry season—this no doubt contributed to
the transport of any red tide that was initiated in Mirs Bay.
A close examination of the circulation in April revealed the
following. A consistent clockwise circulation was formed in the
northeast of Mirs Bay starting from April 5, 1800 hrs the starting
time of the ebb tide. The formation of this distinct clockwise
gyre appears to be related to the clockwise change in wind direc-
tion from northerly winds on April 4 360 N, April 5 50 NE,
and April 7 360 N, to SE winds during April 7–9; this change in
wind direction occurred under almost diurnal tidal conditions.
The longer flooding and ebbing periods provided the chance to
maintain the clockwise circulation. If the wind direction is
changed to 45 NE from April 6 to 12, the clockwise circulation in
north Mirs Bay is still maintained. However, if the wind is
changed to 45 NE from April 4 and maintained in this direction,
the circulation in Mirs Bay is no different from that found in the
typical dry season Fig. 6, with a characteristic counterclockwise
gyre. Likewise, if the actual measured wind direction in 1998 is
shifted backward in time by one week when the tide is semidi-
urnal, no clockwise circulation is formed.
Fig. 13 shows the computed surface current field in the north-
eastern and southern waters for both flood and ebb tides. Based
on the drogue tracking patterns, the path where an initiated bloom
could possibly be transported to the East Lamma Channel is in-
dicated by the gray strip. It seems highly unlikely that the fish
farms in the East Lamma area would be affected by a bloom
originating outside this ‘‘strip of influence.’ Computed velocities
in the East Lamma Channel show that phytoplankton at the sur-
face can be transported in the NW direction along the channel
during a flood tide for an extended period of almost 10 h. Com-
pared to the normal dry season, the surface current near Lamma
Island has increased current speed toward the NW direction more
negative velocitiesup the East Lamma Channel Fig. 9.
It is apparent that any blooms initiated in the northeast of Mirs
Bay during this period would very likely be transported by this
clockwise gyre to the Nan Au area and then into the high velocity
belt ‘‘gray strip’to the southeast of Lamma Island Fig. 13.
The persistent SE wind during the four-day period from April 9
also would then drive the bloom to the East Lamma Channel. This
is illustrated by the close-up view of float tracks released within
the gray strip Fig. 14. Compared to the typical dry season situ-
ation Fig. 14b兲兴, the bloom within the gray strip would more
likely enter the East Lamma Channel under 1998 real wind con-
ditions Fig. 14a兲兴. During a diurnal flood tide, it is highly prob-
able for a bloom inside the gray strip to enter the East Lamma
Channel Lo Tik Wan area兲关Fig. 14c兲兴. However, under a se-
midiurnal spring tide, the bloom would stay for less time inside
the East Lamma Channel Fig. 14d兲兴 and would more likely be
transported to the southeast of Hong Kong Island or to the south
beaches of Hong Kong Island rather than to Lo Tik Wan. This is
consistent with the red tide outbreaks observed in Repulse Bay,
South Bay, and Shek O in mid-April Fig. 2 and Table 2.
Fig. 13. Computed flood and ebb surface currents near Lamma Island during the period of 4/9/98 –4/12/98 path of algal bloom track indicated
by gray strip
Fig. 14. Drogue tracks in Lamma Island area for releases from high
velocity belt in 1998 output interval is 1 day for aand b; 6 h for
cand d兲兴
The algal bloom tracking behavior in the wet season not
shownis distinctly different. In contrast to the dry season where
a bloom initiated in Mirs Bay tends to move from the northeast to
southwest direction, the bloom would tend to move from the
northwest to the southeast direction, away from Hong Kong.
A calibrated 3D hydrodynamic model Delft3Dhas been applied
to study the hydrodynamic transport of red tides in Hong Kong
via extensive numerical drogue tracking experiments. The novel
hydrodynamic tracking of algal blooms has led to valuable in-
sights into algal bloom transport, and unraveled the cause of the
massive 1998 red tide and fish kill observed in Hong Kong’s
southern waters. It is found that, for a typical dry season, a red
tide initiated in the northeastern waters of Hong Kong north Mirs
Bay, Tolo Harboror Daya Bay would not normally be trans-
ported into the East Lamma Channel. On the other hand, the
spring 1998 red tide is very much related to the hydrodynamic
transport. First, the red tide in Kat O in mid-March 1998 may
have originated from the Yiantian or Xiao Mei Sha area, or been
transported from Nan Au by the typical counterclockwise circu-
lation in Mirs Bay. Second, the red tides recorded in Tolo Harbor
in March–April some of which were not toxiccould have origi-
nated from Nan Au or Kat O, or the Tap Mun area. Third, the
massive red tide observed in southern waters and East Lamma
Channel from April 9 to 16 may have originated in Nan Au,
northeast Mirs Bay, or the Tap Mun area. The change in wind
direction under diurnal tide conditions in early April gave rise to
a consistent clockwise gyre in Mirs Bay, which greatly enhanced
Fig. 15. Comparison of model prediction and analytical solution for long wave propagation in rectangular channel: 1Constant depth and 2
quadratic bottom bathymetry
Fig. 16. Comparison of model prediction and analytical solution for long wave propagation in annular channel with quadratic bathymetry: a
Model grid for annular channel and bpredicted and analytic free surface elevation and velocity
the transport of blooms in these areas into the southern coastal
waters and into the East Lamma Channel bordering the Lo Tik
Wan and Sok Kwu Wan fish culture zones. However, it is unlikely
that the red tide in spring 1998 originated from Daya Bay as
speculated by the press.
It is our belief that the advective transport process is, to first
order, the most important. The useful insights gained herein can
aid greatly in disaster mitigation. For example, the general trans-
port pattern under different tidal and wind conditions reported
here can serve as a useful supplement to a real time red tide
warning system. At least, the present findings would contribute
substantially toward the interpretation of future red tide observa-
tions. It is recognized that algal bloom dynamics and other im-
portant factors such as turbulent diffusion, algal growth and sink-
ing, and vertical migration have been neglected; these need to be
addressed in more refined modeling in the future. It is also likely
that the high frequency of red tides observed in 1998 is related to
the water temperature. 1998 was the warmest year in Hong Kong
since records began in 1884; the annual mean air temperature was
24°C compared to a normal 23°C. In January 1998 the water
temperature rose sharply from 20 to nearly 24°C; recent eutrophi-
cation modeling of Tolo Harbor Arega and Lee 2002has shown
that this resulted in more algal growth and nutrient depletion
compared to other years.
This study was supported by a Hong Kong Research Grants
Council Group Research Project RGC/CA/HKU 2/98C. The as-
sistance of Dr. Anthony Lee of the Hong Kong Environmental
Protection Department in the Delft3D Pearl Estuarymodel is
gratefully acknowledged. The writers are indebted to Dr. Patsy
Wong of the Hong Kong Agriculture, Fisheries and Conservation
Department for providing the data of Table 2 and many helpful
discussions. The comments of Dr. Don Anderson and the anony-
mous reviewers on an earlier version of the paper are much
Appendix I. Validation of
The hydrodynamic model adopted for this application, Delft3D,
has been tested for its accuracy; this gives an idea of the Courant
number at which accurate simulations should be performed. In
addition to the analytical tests reported herein, the Pearl River
Estuary model has separately been validated against extensive
tidal, salinity, and velocity data. Three test cases of long wave
propagation are reported below. The open boundary is subjected
to a single harmonic tidal forcing with period T12 h, and the
Courant number is set as 0.24. For each case, a linearized bottom
friction coefficient is used in the analytical solution and the cor-
responding Manning coefficient in the Delft3D hydrodynamic
model which uses a quadratic friction lawcan be obtained by
equating the rate of work done over a tidal cycle. The analytical
solutions for the respective cases can be found in, e.g., Ippen
1966and Lynch and Gray 1978.
Rectangular Channel with Constant Depth
The rectangular channel has a length L5,000 m, width Y
500 m, and mean depth H010 m. 00.5 m is the amplitude
of the tidal forcing at the open boundary. The linear bottom fric-
tion coefficient is 0.0001; the corresponding Manning coefficient
can be shown to be 0.0225. The computed standing wave is com-
pared with the analytical solution in Fig. 15 Case 1.
Rectangular Channel with Quadratic Bathymetry
For this case, the channel depth varies quadratically from9mat
the closed end to 24.95 m at the open end, with an initial bottom
slope of 1/7,500. An average Manning coefficient of 0.0323 is
used, corresponding to a linear bottom friction coefficient of
0.0001. The model predictions of tidal elevation at high tideand
velocity at midtideare compared with the analytical solution in
Fig. 15 Case 2. It is seen that for both cases the prediction is in
excellent agreement with the analytical solution.
Annular Channel with Quadratic Bathymetry
The third test case is the radial flow in an annular section with
outer radius r250,000 m and inner radius r120,000 m. r
5,000 m along a ray direction Fig. 16a兲兴. The depth varies
quadratically as H(r)H03r/2(r2r1)2H0(sloper)2with
Fig. 17. Example comparison of numerical results for a grid with
different layers: aComputed velocity profile near Lamma Island
and bcomputed time history of surface velocity in south of Mirs
Fig. 18. Comparison of computed drogue tracks for grids with
different numbers of vertical layers same release from 4/10/98
LW, 1400 hrs; output interval1 day
H03m,r(r1,r2). The bottom slope parameter from inner to
outer radial directionis 0.00005. Hence the depth varies from 3
m at the closed end to 18.75 m at the open end. Atidal forcing of
0.1 m amplitude is applied at the outer boundary. A Manning
coefficient of 0.0357 is used, corresponding to a linear bottom
friction coefficient of 0.0001. The comparison between the nu-
merical prediction and the analytical solution is shown in Fig.
16b; good agreement is again obtained. It should be noted that
the above serves to validate the accuracy of the numerical solu-
tion of the shallow water equations, excluding the two-equation
turbulence model; the adequacy of the turbulence model for the
vertical structure is supported by validation against salinity data.
Appendix II. Selection of Number of Model Layers
The vertical variation of horizontal velocity computed by using
different vertical layers is compared at key locations e.g., Fig.
17a兲兴. The time variation of surface current for three vertical
model resolutions is also compared for many locations; Fig. 17b
shows the time history of the surface current south of Mirs Bay
inside the gray strip. The results show that the four-layer compu-
tation gives results very close to those of the eight-layer predic-
tion, with the correct phase and occasional minor difference in
peak velocity. On the other hand, based on this result and other
analytical drogue tracking tests in exactly known flow fields, the
two-layer model will likely result in unacceptable errors. As far as
algal bloom tracking is concerned, it is judged that the four-layer
model is both efficient and of sufficient accuracy. Fig. 18 shows a
drogue tracking comparison between eight- and four-layer model
results. In general, the drogue paths are similar, although the
travel time may differ by up to 1–2 days for typical tracking
periods of two weeks.
Anderson, D. M. 1998. ‘‘Study of red tide monitoring and management
in Hong Kong: Literature review and background information.’’Tech-
nical Rep. No. 1, Agriculture and Fisheries Department, Hong Kong
Government, Hong Kong.
Arega, F., and Lee, J. H.W. 2002. ‘‘Long term circulation and eutrophi-
cation model for Tolo Harbour, Hong Kong.’’Water quality and eco-
system modelling, Kluwer, Dordrecht, The Netherlands, Vol. 1, 169–
Chin, P. C., and Leong, H. C. 1978. ‘‘Estimation of wind speeds near
sea-level during tropical cyclone conditions in Hong Kong.’’ Hong
Kong Royal Observatory Technical Note No. 45,http://, 10.
Delft Hydraulics. 1997. ‘‘Upgrading of the water quality and hydraulic
mathematical models.’’ (Draft) Model Validation Rep., Technical Ser-
vices Division, Civil Engineering Office, Hong Kong Government,
Hong Kong.
Delft Hydraulics. 1998. ‘‘Upgrading of the water quality and hydraulic
mathematical models.’’ Final Model Calibration and Validation Rep.,
Technical Services Division, Civil Engineering Office, Hong Kong
Government, Hong Kong.
Dickman, M. D. 1998. ‘‘Hong Kong’s worst red tide.’’Proc., Int. Symp.
on Environmental Hydraulics, J. H. W. Lee et al., eds., Balkema,
Rotterdam, The Netherlands, 641–645.
Hong Kong Observatory. 1998.Monthly Weather Summary February
1998, Hong Kong.
Hong Kong Observatory. 1999.Summary of Meteorological Observa-
tions in Hong Kong (1998), Hong Kong.
Ippen, A. T. 1966.Estuary and coastline hydrodynamics, McGraw-Hill,
New York.
Launder, B. E., and Spalding, D. B. 1974. ‘‘The numerical computation
of turbulent flows.’’Comput. Methods Appl. Mech. Eng., 3, 269–289.
Lee, J. H. W., Wu, R. S. S., and Cheung, Y. K. 1991. ‘‘Forecasting of
dissolved oxygen in marine fish culture zone.’’ J. Environ. Eng.,
1176, 816 833.
Lynch, D. R., and Gray, W. G. 1978. ‘‘Analytic solutions for computer
flow model testing.’’ J. Hydraul. Div., Am. Soc. Civ. Eng., 10410,
Rodi, W. 1980.Turbulence models and their application in
hydraulics—a state of the art review, International Association for
Hydraulic Research, Delft, The Netherlands.
Stelling, G. S. 1984. ‘‘On the construction of computational methods for
shallow water flow problems.’’ Rijkswaterstaat communication, no.
35, Rijkswaterstaat, The Hague, The Netherlands.
Yang, Z. B., Takayama, H., Matsuoka, K., and Hodgkiss, I. J. 2000.
‘‘Karenia digitata sp. no., a new harmful algal bloom species from the
coastal waters of west Japan and Hong Kong.’’ Phycologia, 396,
... Section 4 discusses the jetty-induced changes of momentum terms and 135 their linkage with M2 amplitude change, and examines the sensitivity of M4 tide to tidal 136 forcing, bottom slope and nonlinearity. Finally, a conclusion is given in Section 5. 2 Model setup and validation 151 152 The tidal dynamics responses to sequential harbor constructions are investigated by the 153 process-based model Delft3D, which has been proved to be a successful modeling system for 154 simulating physical processes in coastal environments, in particular in long-term scales (Hu 155 et al., 2009; Lee and Qu, 2004;Liang et al., 2018;van der Wegen and Roelvink, 2012;156 Vousdoukas et al., 2016;Zarzuelo et al., 2015). More importantly, the present tidal model 157 set up by Delft3D-FLOW can be fully two-way coupled with other integrated models, e.g., 158 Delft3D-SED and Delft3D-MOR, to further estimate the impacts of harbor construction on 159 local sediment transport and morphological evolution in near future. ...
Full-text available
The impacts of large-scale coastal structures on tidal dynamics have been studied extensively over the past decades, but the relevant mechanisms are not fully understood. In this study, a two-dimensional Delft3D barotropic tidal model is applied to investigate the tidal dynamic changes induced by the large-scale sequential constructions of Huanghua-Binzhou Harbor in the northeastern China during year 2003–2016. It was found that the predominant M2 tide is increased on the stoss side and decreased on the lee side of the harbor by 2–6 cm due to the harbor constructions. This feature is due to 1) the increased and decreased potential tidal energy on the stoss and lee side, respectively, by the harbor structures; 2) the structure-induced changes in momentum flux is balanced by pressure gradient, and thus leads to a larger tidal range on the stoss side and a smaller tidal range on the lee side. The sensitivity studies on nonlinear M4 tide demonstrate that it is generated locally by M2 self-self interaction, which is not only related to M2 tidal energy but also nonlinearity. Consequently, M4 tide increases towards the shore on the tidal flat with natural slope due to enhanced nonlinearity by decreasing water depth, but is proportional to M2 tide on the tidal flat with a constant water depth, where local nonlinearity is nearly the same in space. In addition, the energy transfer from M2 to M4 is affected by the structure-induced changes in nonlinearity, resulting in different responses of M2 and M4 tide to the coastal structure.
... Comparing with the field observation, numerical modeling is another effective approach used to study water quality problems. Benefiting from the achievement of computer technologies, there have been a number of significant developments in the field of water quality modelling and these have resulted in a variety of commercial packages (Cox, 2003), including QUAL, WASP, MIKE, EFDC, Delft3D (Wool et al., 2003;Lee et al., 2004;Vanderborght et al., 2007;Fan et al., 2009;Lee et al., 2017). ...
Conference Paper
Full-text available
Accidental water pollution incidents, usually in the form of point source, are potential threats to the local drinking water sources. In this study, a two-dimensional water quality model was developed using Delft3D software, aiming to investigate the effect of such accidental incidents on the two drinking water sources in Chengtong Reach, Yangtze River. The model was validated against the measured data, including the water level, the velocity and the phenol concentration in an accidental incident. Four scenarios with different incident times were simulated. The results showed that a large runoff discharge would bring phenol downstream faster, causing a longer propagation path and higher concentration, while the oscillation induced by tide would enhance the dilution but increase the impact duration of accidental incidents. The worst condition for two drinking water sources in the study area is the accidental incident emerging at the spring tide in dry season.
... The constituent transport model is set up based on the Delft3D package, which is an open source model system that consists of several integrated models, e.g., Delft3D-Flow and Delft-WAQ (Delft3D Water Quality) model (Lesser et al., 2004). The Delft3D package has been utilized to in hydrodynamic and water quality simulations around the world (e.g., Bouma et al., 2007;Hartog et al., 2008;Lee and Qu, 2004;Van Dongeren et al., 2008). The constituent transport model contains two parts, the flow model and the transport model. ...
... The algal bloom phenomenon (particularly the red tide) has been widely reported and has become a serious environmental problem due to its adverse influence on aquatic life and on human health. The need for a better understanding of harmful algal bloom (HAB) dynamics and the complex ecological processes involved in blooms is clearly evident from years of research ( Lee and Qu, 2004). In spite of the extensive research that has already been undertaken, the causality and dynamics of algal blooms are not wellunderstood, and the prediction of algal blooms remains a very difficult problem due to the extremely complicated ecological dynamics of these systems (Taranu, GregoryEaves, Steele, Beaulieu & Legendre, 2017). ...
Full-text available
The Great Lakes are most important freshwater bodies providing water resources and other various related businesses to the northeastern part of North America. However, harmful algal blooms (HABs) are more often and severe in those lakes than before and thus threatening lake environments and economies. Researchers have studied the factors influencing HABs characteristics using different scientific methods. In this study, all possible predictors and predictand variables were collected from various data source and then eight final predictors and one predictand were selected based on correlation between predictors and predictand variables. This study tests two machine learning techniques, Stepwise Multiple Regression (SMR) and Genetic Programming (GP), to forecast monthly HAB indicators in Western Lake Erie from July to October. SMR and GP models were created with selected input variables for two training periods, 2002 to 2011 and 2002 to 2014. A Spearman rank correlation coefficient was used to choose input variable sets for each HAB month considering 224 different combinations of lag time and average periods. The SMR models showed a correlation coefficient increase from 0.71 to 0.78 when extending the training period. The GP models followed a similar trend increasing the overall correlation coefficient from 0.82 to 0.96. Both models optimally selected monthly discharge and phosphorus mass from Maumee River Basin as significant predictor variables. A major drawback of both models was data-dependency as common in data-driven methods. GP was better to detect high nonlinear HAB mechanism than SMR due to its nature to use many mathematical functions while SMR only use the linear combination of variables. This study attested that both SMR and GP can be useful to simulate historical HAB event and predict future HAB severity. In future work, to avoid under- or over-prediction for unobserved HAB mechanism regarding short training period, it is suggested to develop an extrapolation technique that is statistically sound and operable in the model and test multi-model ensemble approaches to provide most possible HAB prediction.
... The retentive nature of downwelling in the semi-enclosed coastal systems produces long residence times leading to prolonged periods for cells to grow (Yin, 2003;Glibert et al., 2005;Lai and Yin, 2014). Many large blooms in the eastern waters of Hong Kong were coastal blooms from China that get advected into Hong Kong waters when the tidal cycle (flood tide) and the wind direction form the 'perfect storm' (Lee and Qu, 2004). Therefore, the N. scintillans cells in Port Shelter were probably initially introduced from coastal waters, via the China coastal current (Huang and Qi, 1997), and became concentrated in bays through physical and/or biological processes during Fig. 9. Growth rates (d − 1 ) of Noctiluca scintillans and concentrations of total Chl a (μg L − 1 ) at the initial and final microcosm incubations and in the field during 4 periods [8-10, 14-16 January (one incubation for each station), 22-24 January and 19-21 February (duplicate incubations for each station)]. ...
Noctiluca scintillans, a heterotrophic dinoflagellate responsible for most of the red tides in Hong Kong waters was investigated to determine the influence of biotic and abiotic factors in determining its population dynamics. N. scintillans first occurred in January when temperature was ~ 17 °C, and reached a maximum of 1980 cells L− 1 one month later. N. scintillans interacted with trophic compartments from picoplankton to mesozooplankton, but its population development was largely related to the availability of phytoplankton, especially diatoms. Growth rates in microcosms (with mainly biological factors) ranged from − 0.38 to 1.11 d− 1 and were similar to the field in situ growth rates (physical and biological factors). Thus, optimum hydrographical conditions (temperature and water stability), and a rich food supply were necessary for N. scintillans to bloom. Its spatial distribution pattern was a result of biological and physical coupling, but mainly controlled by physical accumulation processes such as winds, tides and currents.
... A multi-layer perceptron (MLP) neural network with a single hidden layer trained with the back-propagation algorithm [4] with a momentum term was used in this study to model and predict the Chlorophyll fluorescence with lead times from 1-day to 3-days. The hyperbolic activation function was used as the activation function for both hidden and output layers. ...
The occurrence of harmful algal blooms (HAB) in coastal waters is a chronic problem in many parts of the world. The development of real time measurement systems and forecasting models of HAB is important for understanding the algal dynamics and for fisheries and environmental management. In the recent past, data-driven modelling techniques such as artificial neural networks have been increasingly used to model algal bloom dynamics. However the predictability of most of these models is questionable due to the use of interpolated data from low frequency measurements. We present here the first real time modelling and prediction of algal bloom dynamics using artificial neural networks. The model is based on continuous high frequency measurements of water quality and hydro-meteorological parameters at Kat O field monitoring station in Hong Kong. A multi-layer perceptron neural network model trained with back propagation algorithm is employed to model and predict the algal biomass (chlorophyll fluorescence) with different lead times. Daily values of chlorophyll fluorescence, dissolved oxygen, water temperature, solar radiation and wind speed are used as input variables. Multivariate statistical analysis of the input variables and tests with different network scenarios suggest that the use of past values of chlorophyll fluorescence alone is sufficient for algal prediction. Consistent with previous studies, this illustrates the auto-regressive nature of the bloom dynamics in the semi-enclosed coastal waters of Hong Kong. Comparison of predicted and observed chlorophyll fluorescence suggest that the neural network can make reliable real time predictions up to 2-days lead-time.
... The exceptional Karenia mikimotoi bloom in Hong kong waters in 1998 (yin et al. 1999) was related to the 1997/1998 El Niño (widely regarded as the most intense of the twentieth century; Isoguchi et al. 2005 and references cited therein). This bloom is thought to have been advected from more offshore coastal waters into Hong kong waters by easterly winds and a flood tide, and then it continued to develop in various inlets in Hong kong (lee & Qu 2004(lee & Qu , lee et al. 2006. Typical oceanographic conditions in coastal waters of Hong kong result in the waters being well flushed, but in 1998 inshore waters of southern china (including those of Hong kong) were 'trapped' against the coast, and high nutrient concentrations from anthropogenic sources resulted in widespread HAbs. ...
Anthropogenic nutrient enrichment of coastal waters is considered to be one reason for the putative worldwide increase in the occurrence of harmful algal blooms (HABs). This has led to the view that the occurrence of HABs diagnoses the undesirable consequence of anthropogenic nutrient enrichment and thus eutrophication. This view rests on a set of assumptions. In this paper, these assumptions and the arguments that have been made about the increase in HABs and their causes are reviewed. No consensus regarding the role of anthropogenic nutrients in stimulating the occurrence of HABs was found. Attempts to relate trends in HABs to enrichment are confounded by spatial and temporal variability in naturally occurring HABs; the human-mediated transport of HAB species; increased monitoring and reporting; variability in large-scale meteorological forcing; the influence of climate change. For large-biomass HABs, the hypothesis that nutrient enrichment can cause HABs is supported in some water bodies but not in others. The global evidence that enrichment brings about an increase in low-biomass HABs of toxin-producing species is more equivocal. It is concluded that evidence of a link in one coastal region should not be taken as evidence of a general linkage in other coastal regions. The occurrence of HABs and the abundance of HAB species should not be used to diagnose eutrophication unless a link to anthropogenic nutrient enrichment can be demonstrated.
In sub-tropical coastal waters around Hong Kong, algal blooms and red tides are usually first sighted in the Mirs Bay, in the eastern waters of Hong Kong. A calibrated three-dimensional hydrodynamic model for the Pearl River Estuary (Delft3D) has been applied to the study of the physical hydrography of Hong Kong waters and its relationship with algal bloom transport patterns in the dry and wet seasons. The general 3D hydrodynamic circulation and salinity structure in the partially-mixed estuary are presented. Extensive numerical surface drogue tracking experiments are performed for algal blooms that are initiated in the Mirs Bay under different seasonal, wind and tidal conditions. The probability of bloom impact on the Victoria Harbour and nearby urban coastal waters is estimated. The computations show that: i) In the wet season (May-August), algal blooms initiated in the Mirs Bay will move in a clockwise direction out of the bay, and be transported away from Hong Kong due to SW monsoon winds which drive the SW to NE coastal current; ii) In the dry season (November-April), algal blooms initiated in the northeast Mirs Bay will move in an anti-clockwise direction and be carried away into southern waters due to the NE to SW coastal current driven by the NE monsoon winds; the bloom typically flows past the east edge of the Victoria Harbour and nearby waters. Finally, the role of hydrodynamic transport in an important episodic event - the spring 1998 massive red tide - is quantitatively examined. It is shown that the strong NE to E wind during late March to early April, coupled with the diurnal tide at the beginning of April, significantly increased the probability of bloom transport into the Port Shelter and East Lamma Channel, resulting in the massive fish kill. The results provide a basis for risk assessment of harmful algal bloom (HAB) impact on urban coastal waters around the Victoria Habour.
Harmful algal blooms and red tides can lead to beach closures, dissolved oxygen depletion and fish kills, and shellfish contamination. Coastal algal blooms are complex hydro-biological phenomena that are notoriously difficult to predict. We present herein a simple algal bloom forecast model that is derived from deterministic water quality modeling and field observational experience. Algal bloom occurrence is predicted as a function of vertical hydrodynamic stability and nutrient availability. The model requires minimal field measurements as input and is highly suited for operational forecasts. Predictions of the model are compared with field observations. The development of a near real time early warning system for forecasting the likelihood of algal blooms and red tides for fisheries management is presented.
Full-text available
Since the early 1980s excessive discharges of organic waste into Tolo Harbour have created serious problems through nutrient enrichment. A number of eutrophication related problems have been reported which incurred financial losses of billions of dollars and serious ecological imbalance. This work addresses the development, verification and application of a water quality model to synthesize the available large database of water quality and to study water quality management issues of Tolo Harbour. Since detrimental water quality problems usually occur during summer when there is stratification and the water temperature is high, the tidally-averaged hydrodynamics and the relative contribution of gravitational circulation and tidal exchanges in Tolo Harbour have been studied. Gravitational circulation is found to be the dominant mixing process for most of the year, accounting for 70 percent of the mixing. A simple and tractable predictive two-layer mass transport and diagenetic dynamic eutrophication model has been developed. The model computes daily variation of key water quality variables in the water column: algal biomass, dissolved oxygen, organic-nitrogen, ammonium-nitrogen, nitrate-nitrogen, and carboneous oxygen demand. In addition, to study the response and impact of the seabed to the overall eutrophication process a sediment sub-model is developed. The diagenetic sediment sub-model computes explicitly the amount of nutrient recycled and the sediment oxygen demand exerted on the water column. The calibrated model has been validated against a 20 year water quality data base under a wide range of hydro-meteorological and environmental conditions. Both spatial and seasonal variation of observed water quality variables are reproduced. The verified model shows that a significant reduction of total nitrogen loading would be required to meet the water quality objectives, with a recovery time of three months for water column and more than two years for sediment.
Full-text available
The paper reviews the problem of making numerical predictions of turbulent flow. It advocates that computational economy, range of applicability and physical realism are best served at present by turbulence models in which the magnitudes of two turbulence quantities, the turbulence kinetic energy k and its dissipation rate ϵ, are calculated from transport equations solved simultaneously with those governing the mean flow behaviour. The width of applicability of the model is demonstrated by reference to numerical computations of nine substantially different kinds of turbulent flow.
A new unarmoured dinoflagellate, Karenia digitata Yang, Takayama, Matsuoka & Hod,kiss sp. nov., is described on the basis of light and electron microscopy. It has bloomed in Hong Kong and Japan, and has caused significant fish kills and financial loss. High ichthyotoxicity persisted for approximately one week after the blooms had subsided. Karenia digitata cells are 10-26.25 mum long and 10-22.5 mum wide. The epicone is hemispherical or broadly conical; the hypocone is hemispherical and is not concave at the antapex. The girdle displacement varies but is more than 20% of the body length. The sulcus continues into the epicone as small finger-like extensions; it does not reach the antapex. At its centre it is displaced by the girdle. Chloroplasts are numerous, irregular in shape and randomly distributed, though usually clumped. The nucleus is spherical to ovoid and centrally to posteriorly located. The cell has an apical groove, which lies to the right of the sulcal axis, extends in a straight line over the apex to about one-third to half-way down the dorsal epicone, and gradually narrows near its centre. Karenia digitata may be confused with Gymnodinium pulchellum based on size, but the latter has a sigmoid apical groove. Kale,zia digitata shares features with Gyrodinium aureolum, Gymnodinium mikimotoi and Gymnodinium brevisulcatum [recently transferred to the genera Gymnodinium, Karenia and Karenia, respectively, by Daugbjerg et nl. (2000) (Phycologia 39: 302-317)], but differs from them in aspects such as overall appearance, cell thickness, size, the apical groove, sulcus arrangement and the shape and location of the nucleus. It closely resembles Gymnodinium vitiligo and Gymnodinium veneficum (both now transferred to the genus Karlodinium), except that these species are smaller and have a sulcus that extends onto the epicone.
A class of solutions to the linearized shallow water equations is presented. The solutions involve two spatial dimensions plus time, and special emphasis is placed on the dynamic steady state. The results are intended for use in testing and verification of numerical models. The class of problems treated allows an independent assessment of the effects of wind stress, variable bathymetry, frictional dissipation, and nonrectangular boundaries on model performance.
Short-term trends in dissolved oxygen in a weakly flushed marine fish culture zone in Hong Kong are studied by a real-time water-quality model with seven system variables: dissolved oxygen, chlorophyll-a, organic nitrogen, ammonia, nitrate, zooplankton, and sediment algal carbon. The carbon-chlorophyll ratio is computed as a function of a time-variable saturating light intensity. Sediment oxygen demand is related to organic loads from the fish farm as well as algal biomass. Water-quality predictions are compared with well-documented field observations for both background and algal bloom conditions, over time scales on the order of a month. Model simulations demonstrate the importance of light intensity and photosynthetic production in dissolved oxygen variations, which are well predicted. The model also provides useful insights into the causes of severe oxygen depletion and phytoplankton dynamics in a eutrophic subtropical coastal marine ecosystem.
"Textbook compased of lecture notes for a course of the same title organized and held by the hydrodynamics laboratory ofthe departament of civil engineering of the Massachusetts Institute of Technology" Incluye índice