Sea level anomalies in straits of Malacca and Singapore

Applied Ocean Research 58 (2016) 104–117

Contents lists available at ScienceDirect

Applied Ocean Research journal homepage: www.elsevier.com/locate/apor

Sea level anomalies in straits of Malacca and Singapore Serene Hui Xin Tay a,∗ , Alamsyah Kurniawan b , Seng Keat Ooi c , Vladan Babovic a a

Department of Civil and Environmental Engineering, National University of Singapore, 1 Engineering Drive 2, Singapore 117576, Singapore Ocean Engineering Program, Faculty of Civil and Environmental Engineering, Institut Teknologi Bandung, Jl. Ganesha 10, Bandung 40132, Jawa Barat, Indonesia c Tropical Marine Science Institute, National University of Singapore, 14 Kent Ridge Road, Singapore 119223, Singapore b

a r t i c l e

i n f o

Article history: Received 22 September 2015 Received in revised form 16 February 2016 Accepted 2 April 2016 Keywords: Sea level anomalies Hydrodynamic Tide Numerical model Malacca Strait Singapore Strait

a b s t r a c t This paper studies sea level anomaly (SLA) behaviour in Malacca and Singapore straits which serve part of a major maritime trade route between Indian and Pacific Ocean using both observed data and numerical model. Spatio-temporal behaviour of SLA in the region is analyzed based on 15 years of in-situ and remote sensing data. Results show that SLA signatures can be distinctly different in the two straits, with vastly opposite behaviours during certain months. By further analyzing spatial dependency of observed SLA in the region, SLA in Malacca and Singapore straits are found to be under the influence of Indian Ocean and South China Sea, respectively. Based on this insight, a numerical model is built with the appropriate non-tidal forcing derived from meteorological model and satellite dataset to properly represent SLA in Malacca and Singapore straits with Root Mean Square Error of less than 10 cm. With this well calibrated model, the effect of different types of forcing on volume flux through the straits is investigated. Combined tidal and non-tidal forcing in the model gives 4 to 7 × 1011 m3 of annual net westward volume flux through the straits which is four to seven times higher than that of tidal forcing alone. Furthermore with this combined forcing, a distinct seasonal trend with westward net flow during northeast monsoon (November to March) and eastward net flow during southwest monsoon (May to September) can be observed through the straits in the model which is not observed with tidal forcing. The findings of this paper highlight the importance of these non-tidal forcing in the model to obtain accurate SLA and flow representation in the straits that is vital to environmental fate and transport modelling during operational forecast. © 2016 Elsevier Ltd. All rights reserved.

1. Introduction Malacca and Singapore straits serve part of an important maritime trade route between Indian Ocean and Pacific Ocean that links major Asian economies such as India, China, Japan and the Southeast Asian countries. Located between two large water bodies: South China Sea (via Singapore Strait) and Indian Ocean, Malacca Strait is highly complex hydrodynamic system. Indian Ocean is dominated by semi-diurnal tides while South China Sea is dominated by both diurnal and semi-diurnal tides [1]. Tidal waves, especially semi-diurnal M2 tidal component generated at the Indian Ocean [2] and mixed diurnal and semi-diurnal waves from South China Sea meet and interact approximately at southern end of Malacca Strait where it links to Singapore Strait. This also results in a higher tidal range at the west compared to the east within

∗ Corresponding author. E-mail address: [email protected] (S.H.X. Tay). http://dx.doi.org/10.1016/j.apor.2016.04.003 0141-1187/© 2016 Elsevier Ltd. All rights reserved.

Singapore Strait [1]. For example, tidal ranges are between 2.7 m in the west and 1.4 m in the east during spring tide, with 40% of spring tidal range across the strait during neap tide. This creates complicated tidal dynamics along the two narrow straits [3,4]. Fig. 1 illustrates the geography and topography of the region. Although this region is mainly tide-dominated, non-tidal flows in the form of sea level variations are observed in Malacca and Singapore straits [5]. These variations of sea level from deterministic tide are defined as Sea Level Anomalies (SLA) in this paper, and are induced by non-tidal flows related to surface wind, density-driven flow, large scale oceanographic processes, geographical landscape, or abrupt variation of topography. Resultant flow pattern could significantly deviate from the deterministic tidal flow or even cause a flow reversal of tidal flows [6]. SLA observed in Singapore Strait has shown to have characteristics of non-stationary non-periodic ocean behaviour of varying temporal and spatial scales [5]. Tkalich et al. [7] suggest that the local SLA is caused by several large scale oceanographic and meteorological parameters such as winds, atmospheric pressure gradient, sea

S.H.X. Tay et al. / Applied Ocean Research 58 (2016) 104–117

105

Fig. 1. Map showing geography and topography of Malacca and Singapore straits with location of stations and cross sections MS (Malacca Strait) and SS (Singapore Strait).

surface temperature and fresh water run-off. Rao et al. [5] find that SLA in this region is strongly correlated with the monsoon winds of the seasonal monsoon seasons; northeast monsoon from November to February and southwest monsoon from May to August. Especially during northeast monsoon period, strong winds from Asia continent prevails over South China Sea and triggers positive SLA events in Singapore waters [5]. Tkalich et al. [7] conclude that SLA is positive in Singapore Strait and negative at the other end of South China Sea near Taiwan during northeast monsoon, and vice-versa during southwest monsoon. In contrast, there are fewer literatures focusing on SLA in Malacca Strait. A recent SLA study in Malacca Strait by Luu et al. [8] discusses the sea level variability at seasonal, inter-annual and long-term (more than 25 years) time scales along coastline of Malaysia Peninsular based on long term time series observation. Studies by Luu et al. [8] and Soumya et al. [9] have shown that sea level variabilities in Malacca and Singapore straits are affected by coupled ocean-atmosphere oscillations, such ˜ Oscillation (ENSO) and Indian Ocean Dipole as El Nino-Southern (IOD). SLA in Malacca Strait is modulated by ENSO and IOD up to 7 and 5 cm, respectively. Effect of IOD decreases from Andaman Sea towards Singapore Strait [8], while SLA due to ENSO may be up to ˜ ˜ episodes [10]. Nina the order of 5–8 cm during strong El Nino/La Complimenting earlier SLA studies based on analysis of observed data have brought many insights to the behaviour of SLA in Malacca and Singapore straits, process-based numerical models that describe spatial and temporal hydrodynamics can further enhance understanding of SLA, current velocity and volume transport in the region. As this region is tide-dominated, early numerical studies were mostly tidal [11–13] and focusing on Singapore Strait. The so-called Singapore Regional Model built by Kernkamp and Zijl [14] and further improved by Kurniawan et al. [4] is one of the first published numerical models to encompass both Malacca and Singapore straits in its modelling domain. The model is applied in other hydrodynamic studies such as van Maren and Gerritsen [15] and Hasan et al. [16], and model forecast is improved using data-driven techniques [17–23]. Adopting the same open boundary in Malacca Strait, Tay et al. [24] built a numerical model cover-

ing Malacca and Singapore straits and the entire South China Sea and Java Sea with good tidal representation. As for the SLA in this region, most numerical studies [25,26], have been carried out in the barotropic mode as the local non-tidal flows are known to be monsoon-driven rather than baroclinically induced [5,7]. One recent study of non-tidal modelling in this region is Kurniawan et al. [6] who applied basin scale wind over entire South China Sea to represent SLA i.e. wind-induced water level (i.e. surge) in Singapore regional waters covering the vicinity of east coast of Malaysia Peninsular, Singapore Strait and Malacca Strait. Though SLA is well represented in Singapore Strait and east Malaysia Peninsular in their model, SLA in Malacca Strait is poorly represented with tide and wind forcing. SLA along east coast of Malaysia Peninsular has been shown to be caused by basin scale wind blowing over South China Sea [6]. However on the other side of the peninsular, source of SLA in Malacca Strait remains poorly described as local SLA could not be represented using the same wind-driven modelling approach. Results of Kurniawan et al. [6] suggest that SLA in Malacca Strait may originate in large regions located far away—as far as Andaman Sea or Indian Ocean. This motivates need for further understanding of origin of SLA in Malacca Strait and supports need to correctly represent associated mechanisms. Other nontidal hydrodynamic modelling studies solely focusing on Malacca Strait such as Rizal et al. [2] and Chen et al. [27] have focused on the current fields caused by wind during different seasons (monsoons and inter-monsoons). However, other than the monthly surface current vector plots reported by Wyrtki [28], there are no current data available for this area. Therefore no proper quantitative validation of model results was carried out by Rizal et al. [2] and Chen et al. [27]. To the authors’ best knowledge, no numerical modelling study has explicitly shown proper representation of SLA in Malacca Strait. As such, the first objective of this paper is to determine the origin and behaviour of the SLA in Malacca and Singapore straits based on observation data. Based on this knowledge, the second objective of this paper will attempt to properly represent the SLA in numerical model and gain insights to the non-tidal flow in the strait.

106

S.H.X. Tay et al. / Applied Ocean Research 58 (2016) 104–117

Table 1 Geographical location and data availability of five UHSLC stations. UHSLC station

Region located

Latitude (degree)

Longitude (degree)

Data availability (%)

Langkawi Lumut Keling Tanjong Pagar Kuantan

Malacca Strait Malacca Strait Malacca Strait Singapore Strait East coast of Malaysia Peninsular

6.433 4.233 2.217 1.267 3.983

99.767 100.617 102.150 103.850 103.433

99.314 97.119 96.938 92.124 98.344

This paper is organized as follows. Analysis of SLA observation data to understand its behaviour in Malacca Strait is presented in Section 2. Representation of SLA in Malacca Strait using numerical model including modelling approach and results are described in Section 3. Section 4 will conclude and summarize the findings of this paper. 2. Sea level anomalies observation data analysis

Table 2 Root mean square difference and correlation coefficient between UHSLC and DUACS SLA dataset at five stations in year 2004. Station

Root mean square difference (m)

Correlation coefficient

Langkawi Lumut Keling Tanjong Pagar Kuantan

0.074 0.090 0.092 0.141 0.096

0.822 0.652 0.430 0.357 0.818

2.1. Insitu data Bulk of insitu observations used in this paper consist of 15 years (1992–2006) of hourly water level data time series provided by University of Hawaii Sea Level Center (UHSLC) (http://uhslc.soest. hawaii.edu/) at five stations are shown in Fig. 1. Three of these stations; Langkawi, Lumut and Keling are located in Malacca Strait. Stations Tanjong Pagar and Kuantan are located in Singapore Strait and east coast of Malaysia Peninsular, respectively. The latter is included in this paper for the sake of comparison and illustration of SLA behaviour in the region. Table 1 shows geographical locations of these stations and their data availability over the 15 years. Non-tidal water level variation or sea level anomaly is defined as water level which cannot be explained by tidal motion. In this paper, SLA at each of five UHSLC stations is described as residual obtained from subtracting tidal water level from observed water level as: SLA = Observedwaterlevel − Tide

(1)

Tidal constituents have been established using a harmonic analysis of observed water level time series. The set of tidal constituents are optimized using a Fast Fourier analysis of residual and reanalyzed using new set of constituents. Final set consisting of 65 tidal constituents are used to generate tidal level time series at each station. In context of this paper, tidal water level variation has been generated based on the 65 tidal constituents and is referred to as ‘observed tide’ while the residuals determined using Eq. (1) are referred to as ‘observed SLA’. Fig. 2 shows 15 years of observed SLA at two of UHSLC stations; Langkawi and Tanjong Pagar, based on removal of 65 tidal constituents from observed water level data. It is noted that the occurrences of extreme SLA event of these two stations differ from each other. One extreme SLA occurrence is observed at Langkawi during end of 2004 while the two extreme SLA occurrences at Tanjong Pagar are observed during mid-1994 and end of 1999. As Langkawi is located in northern Malacca Strait which is close to Indian Ocean and Tanjong Pagar is located further south in Singapore Strait which is connected to South China Sea, the different extreme SLA occurrences depict different SLA sources at these two stations and weak SLA influence between them. 2.2. Remote sensing data Besides SLA determined from insitu UHSLC stations, another source of SLA data is available through satellite altimetry. These SLA data are produced by Segment Sol Multimission Altimetry and Orbitography (SSALTO)/Developing Use of Altimetry for Climate Studies (DUACS) and distributed by Archiving, Validation and

Interpretation of Satellite Oceanographic Data (AVISO), with support from Centre National d’Etudes Spatiales (CNES) (http://www. aviso.altimetry.fr/duacs/) and is referred to as ‘DUACS SLA’ data in this paper. The DUACS SLA dataset used consists of merged data from available satellites; Saral/AltiKa, Cryosat-2, Jason-1, Jason-2, TOPEX/Poseiden, Envisat, GFO and ERS1/2. Along-track data are mapped to a fixed grid based on optimal interpolation algorithms [29]. In this paper, 14 years (1993–2006) of gridded Delayed-Time SLA data with a 1/4◦ spatial resolution at a daily temporal resolution are used [29]. It should be noted that DUACS SLA data are only available from calendar year 1993 onwards due to the fact that the first satellite altimeter (TOPEX/Poseidon) to providing the first continuous global coverage of sea surface topography was only launched in late 1992. In order to compare quality of UHSLC SLA and DUACS SLA datasets, DUACS SLA values at the five UHSLC station locations have been obtained by linear interpolation of the gridded DUACS SLA maps. Table 2 summarizes associated root mean square difference (RMSD) and correlation coefficient for the year 2004 at all five stations between the two SLA datasets. Low RMSD and high correlation values are observed at stations located closer to the open waters such as Langkawi and Kuantan. Stations in Malacca Strait show decreasing correlation while approaching southwards. Station located in the narrow strait of Singapore such as Tanjong Pagar has higher RMSD and lower correlation values. This indicates that the DUACS SLA data are less reliable when approaching landwards or areas of high landmass density, and application of the SLA data in such areas should be carefully considered. Though the spatial coverage of available SLA data is good in DUACS dataset, its low reliability near the sheltered coastal regions and poor temporal resolution make it unsuitable for model result assessment compared to UHSLC dataset. Nonetheless, the DUACS SLA dataset could serve as a supplement for other model applications in open waters such as open boundary conditions. For model assessment and validation, UHSLC SLA data will be used as observation dataset. 2.3. Overall SLA behaviour Minima, maxima, means and standard deviations of observed SLA at all five UHSLC stations over 15 years are shown in Table 3. Mean SLA at all five stations is zero. In Malacca Strait along west coast of Malaysia Peninsular (Langkawi to Keling), minimum observed SLA values range between −0.865 to −0.669 m, whereas maximum observed SLA values range between 0.431 and 0.750 m. Standard deviation of observed SLA in Malacca Strait is about 0.089–0.117 m and decreases as it proceeds southwards. This could

S.H.X. Tay et al. / Applied Ocean Research 58 (2016) 104–117

107

Fig. 2. 15 years SLA timeseries of two stations Langkawi (top) and Tanjong Pagar (bottom).

Table 3 Minimum, maximum, mean and standard deviation values of observed SLA at 5 UHSLC stations over 15 years (1992–2006). Station

Minimum (m) Maximum (m) Mean (m) Standard deviation (m)

Langkawi Lumut Keling Tanjong Pagar Kuantan

−0.865 −0.835 −0.669 −0.820 −0.498

0.431 0.750 0.438 0.755 0.848

0.000 0.000 0.000 0.000 0.000

0.117 0.107 0.089 0.123 0.160

imply that SLA behaviour in Malacca Strait is influenced by oceanographic effects generated from Andaman Sea or Indian Ocean. In Singapore Strait i.e. Tanjong Pagar, range of observed SLA values is between −0.820 and 0.755 m. Standard deviation of observed SLA at Tanjong Pagar is 0.123 m, which is comparable to that of Malacca Strait but lower than station in east coast of Malaysia Peninsular. It is noted that variance of SLA at Tanjong Pagar is lower than that of Kuantan. This could be due to the fact that Singapore Strait is geographically more sheltered from open sea dynamics compared to Kuantan which faces vast South China Sea. 2.4. Monthly SLA behaviour Figs. 3 and 4 illustrate behavior of means and standard deviations, respectively, for corresponding stations sorted from Langkawi (west coast of Malaysia Peninsular) to Kuantan (east coast of Malaysia Peninsular) on monthly basis. It is observed that stations in west and east coast of Malaysia Peninsular have completely different SLA behaviours. West coast stations experience negative SLA while east coast station (Kuantan) experiences

positive during northeast monsoon (December–March), and vice versa during southwest monsoon (May–September). Signs of SLA on both sides of peninsular are the same during April, October and November which corresponds to usual inter-monsoon period. SLA in Singapore Strait behaves similarly as that of the east coast stations, which is consistent with the findings of Rao and Babovic [30]. In Malacca Strait, mean monthly SLA ranges between −0.15 m and 0.10 m (Fig. 3). Behaviour of SLA over months indicates different hydrodynamic regimes within Malacca Strait, especially when location of stations is considered. The monthly SLA pattern is similar to that reported in Luu et al. [8]. Langkawi located closest to Andaman Sea is directly under influence of Andaman Sea as also captured by closest neighbouring station, Lumut which follows similar monthly SLA behavior. These two stations have significant negative SLA between January and March, positive SLA between May and November, and close to zero in April and December which can be interpreted as transition months, though the SLA is also almost zero in September. In the south of Malacca Strait, Keling is observed not to be under a single source of influence since its SLA behavior follows Langkawi or Tanjong Pagar during certain months with a lower magnitude. This is especially pronounced in December when SLA is highest along east coast facing South China Sea. Spatially- and temporally-varying SLA observed in Malacca Strait based on the above indicate strong (monsoon) seasonal influence from both Andaman Sea and South China Sea, and also extent of influence within Malacca Strait from both seas during different monsoon seasons. Fig. 4 depicts monthly standard deviation organized by stations. In Malacca Strait, standard deviation ranges between 0.060 m and

108

S.H.X. Tay et al. / Applied Ocean Research 58 (2016) 104–117

Fig. 3. Monthly mean observed SLA at five UHSLC stations over 15 years (1992–2006).

Fig. 4. Monthly standard deviation of observed SLA at five UHSLC stations over 15 years (1992–2006).

0.130 m. Monthly standard deviation patterns follow monsoon seasons of the region. Higher standard deviations are observed during northeast monsoon (November–March) and lower standard deviation during southwest Monsoon (May–September) with exception of May. Abnormally high standard deviation at stations Langkawi and Lumut in May could be related to transition of mean SLA from negative to positive influenced by Andaman Sea. Similar to mean, standard deviation shows a similar behavior for different groups of stations dependent on their location in the Malacca Strait. Stations Langkawi and Lumut have similar monthly standard deviation magnitude, while the other three stations have different magnitudes of standard deviation. It is noted that standard deviation of Langkawi is much higher than its neighbour Lumut during months with negative mean SLA (December–April). This indicates strong influence of Andaman Sea during this period. For all months, station Keling has the lowest standard deviation in Malacca Strait. This could be due to its geographical location in the south of Malacca Strait that shelters it from the influence of Andaman Sea. On east coast of Malaysia Peninsular, monthly standard deviations of SLA have distinctly seasonal trend. Higher variance is observed during northeast monsoon than southwest monsoon. SLA standard deviation in Singapore Strait follows this same monthly trend of east coast of Malaysia Peninsular.

2.5. Seasonal spatial SLA pattern Earlier analysis of hourly observed SLA has been carried out using data at sparse local coastal stations. With high temporal resolution, detailed local dynamics and station-to-station relationship can be described and established based on such analysis. However, limited number of observed stations restricts spatial understanding of the SLA behaviour and its dynamics in the region. Using gridded DUACS SLA data, monthly composites of mean and standard deviation of 14 years of SLA data within the model domain have been computed. Four months: January, April, July and October, are selected to represent northeast, first inter-, southwest and second inter-monsoons, respectively. Mean SLA and standard deviation in January, April, July and October are presented in Figs. 5–8, respectively. In January, mean SLA is below zero (around −0.10 m) in northern Malacca Strait, and gradually increases to 0.01–0.05 m toward Singapore. On other side of Malaysia Peninsular, mean SLA is as high as 0.15 m, greater than 0.20 m in Gulf of Thailand, and less than 0.10 m further away from the coast. Mean SLA in the area between Singapore and Borneo is also around 0.15 m. High SLA on eastern side of the Malaysia Peninsular is mostly due to northeast monsoonal wind prevailing over South China Sea during this time

S.H.X. Tay et al. / Applied Ocean Research 58 (2016) 104–117

109

Fig. 5. Mean (left) and standard deviation (right) of DUACS SLA in January from 1993 to 2006.

Fig. 6. Mean (left) and standard deviation (right) of DUACS SLA in April from 1993 to 2006.

Fig. 7. Mean (left) and standard deviation (right) of DUACS SLA in July from 1993 to 2006.

of the year. Standard deviation of the SLA is higher on west coast (0.12–0.14 m) than east coast (around 0.08 m) of Malaysia Peninsular. In open waters, SLA standard deviation ranges between 0.06 and 0.08. In April, mean SLA on both sides of coastal Malaysia Peninsular is around zero, and around −0.05 m in open waters. Standard

deviation of SLA is also smaller than that in January; 0.06–0.12 m in the Malacca Strait, 0.06–0.08 m in east coast of Malaysia Peninsular, and around 0.03–0.05 m in open waters. First inter-monsoon period is a relatively calm season in terms of SLA. In July, mean SLA pattern is opposite of January’s, with negative on east coast of Malaysia Peninsular and positive in Malacca

110

S.H.X. Tay et al. / Applied Ocean Research 58 (2016) 104–117

Fig. 8. Mean (left) and standard deviation (right) of DUACS SLA in October from 1993 to 2006.

Strait which is probably due to switch in prevailing wind direction. Mean SLA in Malacca Strait is about 0.10 m in north and gradually decreases to zero toward south. Mean SLA is about −0.15 m on other side of Malaysia Peninsular and −0.05 to 0 m further out into open waters. Mean SLA in area between Singapore and Borneo ranges between −0.12 and −0.08 m. Standard deviation of SLA is about 0.08–0.12 m in Malacca Strait, 0.04–0.06 m in eastern coast of Malaysia Peninsular, and less than 0.04 m in open waters. In October, mean SLA over Malacca Strait and east coast of Malaysia Peninsular is about the same; around 0.05–0.08 m, and slightly higher (about 0.1 m) in open waters south of Vietnam. However spatial SLA standard deviation shows distinct differences in Malacca Strait (0.10–0.14 m) and east coast of Malaysia Peninsular (0.06–0.08 m). Generally, variance of SLA is perpetually higher in Malacca Strait than in east coast of Malaysia Peninsular over a year. Regardless magnitude, both UHSLC data and DUACS data show similar temporal and spatial SLA patterns in the region. 2.6. Spatial dependencies of SLA To study dependency of SLA at a particular location on its surrounding, correlation coefficient can be computed using SLA established on basis of DUACS dataset from 1993 to 2006. Fig. 9a presents spatial variability of correlation coefficient between SLA at Langkawi and remainder of the region. Correlation is high (more than 0.7) in northern Malacca Strait and in region along coast of Myanmar to Bangladesh. The area west of Sumatra shows correlation of greater than 0.6. This implies that SLA in northern Malacca Strait is associated with SLA originating in east Indian Ocean and both most probably generated by the same source. It is noted that this strong correlation terminates at Singapore Strait and western Bay of Bengal. Weak negative correlation (around −0.1) is observed on other side of Malaysia Peninsular. At the southern part of Malacca Strait, local SLA dependencies are illustrated using correlation coefficient between SLA at Keling and remainder of the region and are illustrated in Fig. 9b. Correlation pattern is similar to that of Langkawi in Malacca Strait and in region along coast of Myanmar to Bangladesh, except slightly weaker (about 0.5). However, there is slightly higher correlation (about 0.4) between SLA at Keling and SLA in Singapore Strait and east coast of Malaysia Peninsular, as compared to that of Langkawi which is almost zero in this region. This illustrates the weakening of east Indian Ocean SLA as it approaches further south in Malacca Strait. Fig. 9c illustrates correlation coefficient of SLA at Tanjong Pagar with SLA at other areas in the region. Correlation is more than 0.8

in area around Singapore Strait, and reduces to 0.65 in shelf area of South China Sea including east coast of Peninsular Malaysia, Gulf of Thailand and Gulf of Tongking. Correlation in Malacca Strait reduces sharply at south of Malacca Strait near Keling. This indicates that SLA of Andaman Sea via Malacca Strait has little influence on SLA in Singapore Strait, and vice-versa. Fig. 9d illustrates correlation coefficient of SLA at Kuantan and SLA of other areas in the region. Similar to spatial correlation of Tanjong Pagar, spatial correlation of Kuantan shows the same spatial pattern (high correlation in shelf area of South China Sea and northern Java Sea) but with higher correlation values (close to 1). This implies that SLA at both Kuantan and Tanjong Pagar are generated by the same source which probably originates in South China Sea. It is noted that there is a strong negative correlation (around −0.8) in deep northern South China Sea near Luzon Strait which suggests possible quasi basin seiching effect in South China Sea, and consistent with the finding reported by Rao and Babovic [30] and Tkalich et al. [7]. Analyses of SLA observation from both in-situ and remote sensing datasets carried out here have shown that SLA in Malacca Strait is dependent on SLA originating from east Indian Ocean instead of SLA from South China Sea or Java Sea. This also explains good SLA representation in Singapore Strait and east coast of Malaysia Peninsular attained by Ooi et al. [26] and Kurniawan et al. [6], as the dominant SLA source of these two areas is resolved by a model driven with basin-scale wind over South China Sea. 3. Numerical modelling Statistical analysis of observed SLA presented in earlier chapter brought insights to monthly behavior of SLA in different areas of Singapore Strait and coastal Malaysia Peninsular. The present section will attempt to represent the SLA of year 2004 in Malacca and Singapore straits using numerical model. Depth-integrated hydrodynamic modelling approach applied by Kurniawan et al. [6] has shown that barotropic model driven by tide and wind forcing can represent tidal and non-tidal barotopic effects (surge and tidesurge interaction) i.e. SLA well, especially in Singapore Strait and east coast of Malaysia Peninsular. In their study, they applied a multi-scale modelling to simulate non-tidal barotropic effect originated from South China Sea basin by driving the models with both tidal and wind forcing. This approach involves offline nesting (oneway coupling) of a smaller and finer resolution model (focusing on Singapore Strait and coastal Malaysia Peninsular) in South China Sea basin scale model. Building on this knowledge, this paper utilizes a model whose domain covers entire South China Sea and applies a multi-domain

S.H.X. Tay et al. / Applied Ocean Research 58 (2016) 104–117

111

Fig. 9. Correlation of SLA between reference point: Langkawi (a), Keling (b), Tanjong Pagar (c) and Kuantan (d) and other areas in the region from 1993 to 2006.

modelling approach i.e. domain decomposition (two-way coupling) where two or more model domains of different spatial resolutions covering coast of Singapore and Malaysia Peninsular and South China Sea are used to simulate both tide and SLA. Unlike one-way coupling technique, two-way coupling technique involves both large and small spatial domain models to compute hydrodynamic conditions simultaneously and communicate with each other at every time step. This in turn reduces computational cost and eliminates the hassle of simulating two models i.e. basin model and smaller finer model, one after another. The numerical model used in this paper is built in the open source Delft3D software environment [31,32]. The multi-domain model is constructed based on single domain South China Sea model named ‘South China Sea Model Curvilinear (SCSMC)’ built by Tay et al. [24]. Multi-domain modelling of SCSMC begins with division of model domains into two sub-domains of different spatial resolutions as shown in Fig. 10. In context of this paper, the model is referred to as ‘SCSMC domain decomposition 2’ (SCSMCdd2) with number ‘2’ referring as two-domain modelling approach. The fine grid domain covers southern South China Sea including Singapore and Malacca straits at spatial resolution of about 2–3 km, while the coarse grid domain adopts the same grid resolution of the original SCSMC (about 5–40 km), with 39,235 grid cells. Bathymetry data is based on ETOPO1 dataset [33] whereas bathymetry of other numerical models covering the same domain such as Gerritsen et al. [34] and Kurniawan et al. [4]. Bed roughness represented by Manning friction coefficient spatially varies from 0.015 to 0.400 m1/3 /s over

the model domain. Time step of model is 5 min, and it takes about 3.8 h computational time for 1 year simulation on an Intel Core i7-2600 (quad core) 3.4 GHz CPU PC. 3.1. Definition of computed SLA In context of this paper, where barotropic effect is of focus, water level variation is considered to consist of net effect of tide and surge. As Kurniawan et al. [6] highlighted significance of non-linear tide surge interaction induced by change of wave speed and dissipation rate over shallow waters [35], water level component of tide surge interaction could not be ignored and is included as part of the nontidal barotropic water level. Therefore following the methodology proposed by Kurniawan et al. [6], total barotropic water level is expressed as sum of tide, surge and tide surge interaction. Since surge and tide surge interaction constitute components of non-tidal water level, their net effect could also be comprehended as SLA. To quantify computed SLA in numerical model, two types of model simulation are made: (1) one with tidal forcing only, and (2) one with both tidal and non-tidal forcing. Computed SLA (SLAcomputed ) is then defined and quantified by subtracting water level of the former (Tidecomputed ) from that of the latter (WLcomputed(Tide, Non-tidal forcing) ), as defined below. SLAcomputed = WL computed(Tide,Non-tidalforcing) − Tidecomputed

(2)

Therefore, in order to have a good SLA representation in the model, a good tidal representation is a pre-requisite.

112

S.H.X. Tay et al. / Applied Ocean Research 58 (2016) 104–117

Fig. 10. Grid schematization of multi-domain model SCSMCdd2: coarse grid domain (Blue) and fine grid domain (Red) of overall model domain (left) and Malacca and Singapore straits (right). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

3.2. Tidal forcing

3.3. Non-tidal forcing

Tidal forcing implemented on the open boundaries is prescribed as water level by means of amplitudes and phases (relative to time zone GMT + 8) of the eight main tidal constituents; O1 , K1 , M2 , S2 , Q1 , P1 , N2 and K2 . It is noted that open boundary tidal forcing applied in SCSMCdd2 is the same as SCSMC [24]. Tidal representation is assessed using the differences between modelled and observed tidal amplitudes and phases of available tidal constituents at observed stations denoted as s which are then computed as the vector difference (VD) [36]:

In addition to tidal forcing, numerical model is driven with nontidal forcing to represent both tide and SLA. In barotropic modelling approach, meteorological data in the form of wind and atmospheric pressure is one of the commonly applied surface boundary condition to generate non-tidal barotropic flows and SLA. Meteorological data used for driving hydrodynamic model are reanalyzed global meteorological data obtained from ECMWF ERA-Interim database (www.ecmwf.int). Since the present modeling approach focuses on barotropic oceanographic effects, only wind and atmospheric pressure fields are used, and resultant water level is referred to as surge. Meteorological data are provided in 0.75◦ spatial resolution with a 6 h temporal resolution. Based on spatial dependencies analysis of SLA using long term DUACS SLA data, resulting strong correlation between SLA in Malacca Strait and SLA in east Indian Ocean suggests that SLA in Malacca Strait originates from Indian Ocean, rather than being solely local wind-induced. Despite this conjecture it remains infeasible and computationally prohibitive to model SLA in the entire Indian Ocean to simulate SLA in Malacca Strait. Therefore to account for this SLA that originates beyond the model domain, this paper proposes to impose SLA directly at Andaman Sea open boundary of the model (Fig. 11). In other words, in addition to tidal water level prescription at Andaman Sea open boundary, SLA water level is applied as open boundary forcing. This additional SLA forcing, referred to as ‘tilt’ [38], is in form of water level time series obtained from DUACS SLA dataset. Daily DUACS SLA values (tilt) at location of Andaman Sea open boundary support points are obtained by linear interpolation of gridded DUACS SLA maps and presented in Fig. 11.

VDk,s =



2 (Ac,k cosGc,k − Ao,k cosGo,k ) + (Ac,k sinGc,k − Ao,k sinGo,k )

2

(3)

With Ac,k , Gc,k , Ao,k and Go,k denoting computed (c) and observed (o) astronomical amplitudes and phases of a tidal constituent k. In other words, a lower VD value will indicate a better tidal representation computed by the model. In this paper, eight tidal constituents (O1 , K1 , M2 , S2 , Q1 , P1 , N2 and K2 ) are considered in the tidal representation assessment. At each station, VD of eight tidal constituents are summed up with the final value referred as summed vector difference (SVD). Observed amplitudes and phases of eight tidal constituents are determined using harmonic analysis based on UHSLC water level dataset. Though not presented in this paper, it is noted that overall tidal representation of the entire model domain of SCSMCdd2 has been assessed based on SVD of 119 altimetry track stations covering the whole model domain mainly in open seas. Amplitude and phase of eight principal tidal constituents of these altimetry track stations are derived from 15 years of TOPEX/Poseidon and Jason-1 altimeter data from RADS database [37]. Overall tidal representation of SCSMCdd2 is comparable to that of SCSMC [24]. As for evaluation of tidal representation in the coastal region such as Malacca Strait, SVD of eight tidal constituents of SCSMCdd2 are determined at five UHSLC stations (Fig. 1) and compared to two other models: SCSMC and a well calibrated localized tidal model Singapore Regional Model (SRM) [4] (Table 4). SCSMCdd2 gives the best tidal representation among the three at the five UHSLC stations except at Langkawi and Kuantan. Nonetheless, the sum of SVD at five stations show that SCSMCdd2 provides the best tidal representation when these stations are considered.

3.4. Model results The following presents result of the model especially the SLA representation using different configurations of non-tidal forcing i.e. surface wind and pressure, and DUACS SLA tilt. The first part focuses on validation of model by comparing computed and observed SLA. With SLA properly represented in the model, volume transport through Malacca and Singapore straits due to different types of forcing is examined.

S.H.X. Tay et al. / Applied Ocean Research 58 (2016) 104–117

113

Table 4 Summed vector difference of eight tidal constituents at five stations simulated by three numerical models. UHSLC station

Langkawi Lumut Keling Tanjong Pagar Kuantan Sum

Models South China Sea Model Curvilinear [24]

Singapore Regional Model [4]

SCSMCdd2 (presented in this paper)

0.319 0.384 0.544 0.950 0.267 2.464

0.197 0.344 0.452 0.491 0.442 1.926

0.376 0.168 0.314 0.406 0.324 1.588

Fig. 11. Andaman Sea open boundary support points (left) and corresponding DUACS SLA tilt (m) (right).

Table 5 Root mean squared error and correlation coefficient of computed SLA using different non-tidal forcing. Non-tidal forcing

Surface wind and pressure forcing

Surface wind and pressure with DUACS tilt forcing

Station

RMSE (m)

Correlation coefficient

RMSE (m)

Correlation coefficient

Langkawi Lumut Keling Tanjong Pagar Kuantan

0.152 0.145 0.129 0.088 0.094

0.405 0.164 −0.121 0.767 0.834

0.098 0.095 0.096 0.087 0.094

0.850 0.829 0.549 0.761 0.834

3.4.1. SLA representation In the assessment of model prediction, root mean square error (RMSE) and correlation coefficient between the hourly UHSLC SLA observation and model’s SLA prediction are computed using different non-tidal forcing configurations for calendar year 2004 and presented in Table 5. Lower RMSE value and higher positive correlation coefficient indicate a better model prediction of SLA. With only surface wind and pressure as non-tidal forcing, model represents SLA better in Singapore Strait and east coast of Malaysia Peninsular compared to Malacca Strait. RMSE of stations from Tanjong Pagar and Kuantan are 0.088 and 0.094 m, respectively, while RMSE of stations from Langkawi to Keling are in range between 0.129 and 0.152 m. Based on correlation coefficient values, difference in quality of SLA representation between east and west coasts of Malaysia Peninsular is more apparent. Computed SLA at east coast is highly correlated to observed SLA with correlation value of 0.834, and 0.767 in Singapore Strait, while poor SLA correlation in Malacca Strait with correlation coefficient values ranging between −0.121 and 0.405. However, with additional non-tidal forcing of DUACS tilt at Andaman Sea open boundary, SLA representation in Malacca Strait improves significantly in terms of RMSE and correlation. RMSE of SLA in Malacca Strait is reduced to less than 0.100, with significant correlation improvement, especially at Langkawi and Lumut

(more than 0.800). Though SLA representation in Malacca Strait has improved significantly by applying DUACS tilt, correlation at the southern end (where Keling is located) remains relatively low (0.549). It is noted that this location is characterized by tidal mixing between Andaman Sea and South China Sea, and may act as a blockage for SLA propagation [6]. At Tanjong Pagar and Kuantan, RMSE and correlation of SLA computed using additional non-tidal forcing of DUACS tilt do not differ much from the result of model without tilt. Effects of the tilt are small in Singapore Strait and negligible on east coast of Malaysia Peninsular. Figs. 12–14 compare computed SLA time series of observation and model using different non-tidal forcing at Langkawi, Tanjong Pagar and Kuantan in calendar year 2004. It is observed that SLA representation of surface wind and pressure driven model at Langkawi fluctuate along −0.100 m water level, while observed SLA varies between −0.400 and 0.400 m over the year (Fig. 12). This shows that driving model with just local surface wind and pressure will not represent SLA in Malacca Strait. Model driven by combined surface wind and pressure and DUACS SLA tilt, on the other hand, computes SLA with trend similar to observation, though the period from April to October is slightly underestimated. This illustrates that DUACS SLA tilt forcing improves SLA representation at Langkawi significantly in the model. At Tanjong Pagar (Fig. 13) and Kuantan (Fig. 14), very small difference in computed SLA

114

S.H.X. Tay et al. / Applied Ocean Research 58 (2016) 104–117

Fig. 12. SLA at Langkawi observed and modelled using non-tidal forcing (1) surface wind and pressure, and (2) combined surface wind and pressure and DUACS tilt.

Fig. 13. SLA at Tanjong Pagar observed and modelled using non-tidal forcing (1) surface wind and pressure, and (2) combined surface wind and pressure and DUACS tilt.

Fig. 14. SLA at Kuantan observed and modelled using non-tidal forcing (1) surface wind and pressure, and (2) combined surface wind and pressure and DUACS tilt.

between models driven by the two non-tidal forcing configurations is observed. At Tanjong Pagar, slightly lower SLA computed without DUACS tilt is observed during April, May, October and November which are the inter-monsoon months. Kuantan, on the other hand, shows no difference between the two computed SLA. This illustrates that effect of DUACS SLA tilt on the other side of Malaysia Peninsular is negligible beyond Singapore Strait. With either nonforcing, computed SLA represents observed SLA well, especially three positive SLA peak events in first quarter of the year. Furthermore, seasonal positive and negative SLA trend during northeast and southwest monsoons, respectively, is also well represented

in models. Such computed SLA result (good SLA representation in east coast of Malaysia Peninsular and poor SLA representation in Malacca Strait) is consistent with Kurniawan et al. [6] who apply a similar barotropic modelling approach with surface wind and pressure as the sole non-tidal forcing. Analysis of observed SLA in earlier chapter has shown that SLA in west and east coasts of Malaysia Peninsular are in different non-tidal regimes, and SLA in Singapore Strait resembles more of SLA of east coast. Model results here reinforce the idea that SLA observed in Malacca Strait is not caused by local barotopic effects or SLA generated in South China Sea.

S.H.X. Tay et al. / Applied Ocean Research 58 (2016) 104–117

115

Fig. 15. Cumulative volume flux through cross-section MS (Malacca Strait) computed using different forcing over year 2004 (negative indicates a westward/northward progression).

Fig. 16. Cumulative volume flux through cross-section SS (Singapore Strait) computed using different forcing over year 2004 (negative indicates a westward/northward progression).

3.4.2. Volume flux transport across Malacca and Singapore straits Based on well represented SLA in Malacca Strait and Singapore Strait by numerical model shown earlier, hydrodynamics computed by the model can be considered to be fairly accurate. To study the effect of different types of forcing on hydrodynamics in these two straits, volume flux across cross-section MS (Malacca Strait) and SS (Singapore Strait) (Fig. 1) modelled using different forcing configurations will be compared. These model forcing include (1) tide only, (2) combined tide and surface wind and pressure forcing, and (3) combined tide, tilt and surface wind and pressure forcing. The latter two model forcing configurations are regarded as tidal and non-tidal forcing. Figs. 15 and 16 present cumulative volume flux through Malacca Strait (MS) and Singapore Strait (SS) (Fig. 1) computed by model driven by the three model forcing configurations over entire year 2004. With application of non-tidal forcing computed volume flux through both straits significantly increases westwards. A distinct seasonal trend with westward net flow during northeast monsoon (November–March) and eastward net flow during southwest monsoon (May–September) can be observed. Compared to model driven by both tidal and non-tidal forcing, tidally driven model presents four to seven times lower net volume flux through the straits within one year. Since width and area of both cross sections differ greatly, the difference in volume flux is almost double. Nonetheless, both straits show a similar seasonal pattern in volume flux exchange. Comparing result of model driven by tide, tilt

and wind and model driven by tide and wind, it can be deduced that tilt induces higher westward flux progression during January to February, and eastward flux progression during May to July, indicated by steeper gradient of rising and falling of net volume flux. This corresponds well to minimum and maximum of applied DUACS SLA tilt (Fig. 11). Despite the fact that SLA computed by these two types of non-tidal forcing may seem to be deviating trivially over a year, differences in net volume flux is significant. It is noted that flow reversal of volume flux through the straits is brought forward by a month; occurring in May instead of June, due to the tilt. Tilt has also reduces annual net westward flow of volume flux by about 5 × 1011 m3 and 3 × 1011 m3 through Malacca and Singapore straits, respectively.

4. Summary and conclusions In this paper, 15 years of observed SLA in Singapore Strait and coastal Malaysia Peninsular have been analyzed to study spatial and temporal SLA behaviour in the region. It was found that SLA behaviour can be classified into two zones: east and west coasts of Malaysia Peninsular; with Singapore Strait behaving more like east coast. Monthly classification of SLA showed a distinct seasonal SLA variation of the two zones. SLA behaviour in Malacca Strait (west coast of Malaysia Peninsular) can be categorized into three temporal regimes: negative SLA between January and March, positive SLA between May and November, and close to zero in

116

S.H.X. Tay et al. / Applied Ocean Research 58 (2016) 104–117

April and December which are the transition months. On east coast of Malaysia Peninsular, temporal behaviour of SLA correlates well with monsoons: positive SLA during northeast monsoon (May–September) and negative SLA during southwest monsoon (November–March). Maps of SLA correlation based on gridded satellite dataset show spatial dependencies of SLA in the region. Results confirm that the SLA in Malacca Strait and Singapore Strait originates from SLA in northeastern Indian Ocean and South China Sea, respectively. Earlier and present studies have shown that the barotropic modelling approach proposed by Kurniawan et al. [6] (i.e. driven by tide and surface wind and pressure forcing) represents SLA well in east coast of Malaysia Peninsular and Singapore Strait, but not in Malacca Strait. Based on the results of both modelling and observed SLA analysis, SLA observed in Malacca Strait is not caused by local barotopic effects or SLA generated in South China Sea, and is generated from a larger oceanographic system of Indian Ocean. To address this phenomenon in the numerical model, this paper imposes SLA or tilt directly on the open boundary adjacent to Malacca Strait at Andaman Sea open boundary. With this additional non-tidal forcing of DUACS tilt, the model gives better overall SLA representation in Malacca Strait with more than 20% improvement than that of the previous modelling approach. With this well-calibrated numerical model, effects of non-tidal forcing i.e. wind, pressure field and tilt on hydrodynamics of Malacca and Singapore straits are investigated. Besides water level, non-tidal forcing has been also observed to cause deviation of volume fluxes through the straits. There are periods (March–May and September–October) during which occurrence of flow reversal due to non-tidal forcing is observed. Volume flux is directed westward during northeast monsoon and eastward during southwest monsoon. Net volume flux through the straits over a year is about 5 to 7 × 1011 m3 and is directed towards west, into Andaman Sea. It is noted that northern Malacca Strait is located close to the model open boundary in the Andaman Sea. This makes the region to be more susceptible to boundary effect and uncertainties due to boundary forcing of tide and DUACS tilt. Therefore to obtain more accurate volume flux estimation in northern Malacca Strait, extension of model domain that includes Bay of Bengal is recommended. As a whole, driving the numerical tidal model with wind, pressure and DUACS tilt is recommended to provide accurate SLA representation in Malacca and Singapore straits. This non-tidal forcing plays a significant role in the net annual and seasonal volume flux transport in this region that is important for studies related to environmental fate and transport modelling during operational forecast.

Acknowledgements The authors gratefully acknowledge the support and contributions of the Singapore–Delft Water Alliance (SDWA). The research presented in this work was carried out as part of the SDWA’s “MustHave Box” research program (R-264-001-003-272).

References [1] M. Chen, K. Murali, B.C. Khoo, J. Lou, K. Kumar, Circulation modelling in the strait of Singapore, J Coast Res 21 (2005) 960–972. [2] S. Rizal, P. Damm, M.A. Wahid, J. Sündermann, Y. Ilhamsyah, T. Iskandar, et al., General circulation in the Malacca Strait and Andaman Sea: a numerical model study, Am J Environ Sci 8 (2012) 479–488. [3] E.S. Chan, P. Tkalich, K.Y.H. Gin, J.P. Obbard, The physical oceanography of Singapore coastal waters and its implications for oil spills. The environment in Asia Pacific Harbours, Springer, 2006, p. 393–412. [4] A. Kurniawan, S.K. Ooi, S. Hummel, H. Gerritsen, Sensitivity analysis of the tidal representation in Singapore Regional Waters in a data assimilation environment, Ocean Dyn 61 (2011) 1121–1136.

[5] R. Rao, H. Gerritsen, H.F.P. van den Boogaard, V. Babovic, South China Sea winds as a triggering mechanism for sea level anomalies around the Singapore coast, in: 9th hydroinformatics conference, Tianjin, China, 2010. [6] A. Kurniawan, S.H.X. Tay, S.K. Ooi, V. Babovic, H. Gerritsen, Analyzing the physics of non-tidal barotropic sea level anomaly events using multi-scale numerical modelling in Singapore regional waters, J Hydro Environ Res 9 (2015) 404–419. [7] P. Tkalich, P. Vethamony, M.T. Babu, P. Malanotte-Rizzoli, Storm surges in the Singapore Strait due to winds in the South China Sea, Nat Hazards 66 (2012) 1345–1362. [8] Q. Luu, P. Tkalich, T. Tay, Sea level trend and variability around Peninsular Malaysia, Ocean Sci 11 (2015) 617–628. [9] M. Soumya, P. Vethamony, P. Tkalich, Inter-annual sea level variability in the southern South China Sea, Global Planet Change 133 (2015) 17–26. [10] P. Tkalich, P. Vethamony, Q.-H. Luu, M. Babu, Sea level trend and variability in the Singapore Strait, Ocean Sci 9 (2013) 293–300. [11] H.-F. Cheong, M.H. Abdul Khader, C.-J. Yang, On the numerical modelling of tidal motion in Singapore Strait, J Hydraul Res 29 (1991) 229–242. [12] C.-T. Chan, H.-F. Cheong, N.J. Shankar, A boundary-fitted grid model for tidal motions: orthogonal coordinates generation in 2-D embodying Singapore coastal waters, Comput Fluids 23 (1994) 881–893. [13] N.J. Shankar, H.F. Cheong, C.T. Chan, Boundary fitted grid models for tidal motions in Singapore coastal waters, J Hydraul Res 35 (1997) 3–19. [14] H.W.J. Kernkamp, F. Zijl, Further hydraulic model studies for Pulau Ubin & Pulau Tekong reclamation scheme, Delft Hydraulics Report Z3437 for Housing and Development Board (HDB), 2004. [15] D.S. van Maren, H. Gerritsen, Residual flow and tidal asymmetry in the Singapore Strait, with implications for resuspension and residual transport of sediment, J Geophys Res 117 (2012) C04021. [16] G.M.J. Hasan, D.S. van Maren, S.K. Ooi, Hydrodynamic modelling of Singapore’s coastal waters: nesting and model accuracy, Ocean Model 97 (2016) 141–151. [17] S.A. Sannasiraj, H. Zhang, V. Babovic, E.S. Chan, Enhancing tidal prediction accuracy in a deterministic model using chaos theory, Adv Water Resour 27 (2004) 761–772. [18] V. Babovic, S.A. Sannasiraj, E.S. Chan, Error correction of a predictive ocean wave model using local model approximation, J Mar Syst 53 (2005) 1–17. [19] Y. Sun, P. Sisomphon, V. Babovic, E.S. Chan, Applying local model approach for tidal prediction in a deterministic model, Int J Numer Methods Fluids 60 (2009) 651–667. [20] Y. Sun, V. Babovic, E.S. Chan, Multi-step-ahead model error prediction using time-delay neural networks combined with chaos theory, J Hydrol 395 (2010) 109–116. [21] R.R. Karri, A. Badwe, X. Wang, G. El Serafy, J. Sumihar, V. Babovic, et al., Application of data assimilation for improving forecast of water levels and residual currents in Singapore regional waters, Ocean Dyn 63 (2013) 43–61. [22] R.R. Karri, X. Wang, H. Gerritsen, Ensemble based prediction of water levels and residual currents in Singapore regional waters for operational forecasting, Environ Model Softw 54 (2014) 24–38. [23] X. Wang, V. Babovic, Enhancing water level prediction through model residual correction based on Chaos theory and Kriging, Int J Numer Methods Fluids 75 (2014) 42–62. [24] S.H.X. Tay, A. Kurniawan, S.K. Ooi, V. Babovic, Further improvement of tidal representation in the South China Sea and the Southeast Asian waters, in: 2013 IAHR congress, Chengdu, 2013. [25] S.K. Ooi, P. Zemskyy, P. Sisomphon, H. Gerritsen, D.J. Twigt, The effect of grid resolution and weather forcing on hydrodynamic modelling of South East Asian waters, in: 33rd IAHR congress, Canada, 2009. [26] S.K. Ooi, P. Sisomphon, A. Kurniawan, H. Gerritsen, Modeling of sea level and current anomalies in the Singapore region, in: IAHR 2011 congress, Brisbane, 2011. [27] H. Chen, P. Malanotte-Rizzoli, T.Y. Koh, G. Song, The relative importance of the wind-driven and tidal circulations in Malacca Strait, Cont Shelf Res 88 (2014) 92–102. [28] K. Wyrtki, Physical oceanography of the Southeast Asian waters, University of California, Scripps Institution of Oceanography, La Jolla, Calif, 1961. [29] AVISO. SSALTO/DUACS user handbook: (M)SLA and (M)ADT Near-Real TIme and Delayed Time Products. Rep. CLS-DOS-NT-06-034; 2014. [30] R. Rao, V. Babovic, Establishing sea level anomaly patterns using mutual information theory, in: 33rd IAHR Congress, Vancouver, Canada, 2009. [31] Deltares, Delft3D-FLOW. Simulation of multi-dimensional hydrodynamic flows and transport phenomena, including sediments. User Manual, Version 3.15. 2011. [32] G.R. Lesser, J.A. Roelvink, J.A.T.M. van Kester, G.S. Stelling, Development and validation of a three-dimensional morphological model, Coast Eng 51 (2004) 883–915. [33] Amante C. Eakins BW. ETOPO1 1 Arc-Minute Global Relief Model: Procedures, Data Sources and Analysis. NOAA Technical Memorandum NESDIS NGDC-24. 2009:19 pp. [34] H. Gerritsen, E.J.O. Schrama, H.F.P. van den Boogaard, Tidal model validation of the seas of South-East Asia using altimeter data and adjoint modelling, in: XXX IAHR congress, Thessaloniki, 2003, pp. 239–246. [35] D. Idier, F. Dumas, H. Muller, Tide-surge interaction in the English channel, Nat Hazards Earth Syst Sci 12 (2012) 3709–3718. [36] C. Le Provost, M.-L. Genco, F. Lyard, Modeling and predicting tides over the world ocean, in: D.R. Lynch, A.M. Davies (Eds.), Quantitative skill assessment

S.H.X. Tay et al. / Applied Ocean Research 58 (2016) 104–117 for coastal ocean models, American Geophysical Union, Washington, DC, 1995. [37] M. Naeije, E.J.O. Schrama, E. Doornbos, R. Scharroo, The role of RADS in building the 15-year altimeter record, ESA Workshop, Venice, 2006, p. 6. [38] H. Gerritsen, R. Uittenbogaard, E.J. Schrama, Modelling of monsoon driven flows in South-East Asian waters using satellite remote sensing techniques,

in: S.Y. Liong, K.K. Phoon, V. Babovic (Eds.), Proceedings of the 6th International Conference on Hydroinformatics, Singapore, 2004, pp. 1285–1294.

117