Applied Mathematical Sciences, Vol. 7, 2013, no. 16, 751 - 764

Statistical Analysis of Sea Surface Elevation in Numerical Ocean Model for the Gulf of during

N. Aschariyaphotha, S. Klanklaew, B. Wichianchai

and S. Wanchaijiraboon

Department of Mathematics King Mongkut’s University of Technology Thonburi 10140, Thailand

Abstract

In this research the Princeton Ocean Model (POM) is used to study the sea surface elevation (SSE) during typhoon that moved over the (GoT), covering the domain 6◦N to 14◦N and 99◦E to 105◦E. The simulation concerns a case of Typhoon Muifa which oc- curred during November 21-28, 2004. The model is the time-dependent, primitive equation, Cartesian coordinates in a horizontal and sigma co- ordinate in the vertical. The model grid has 37 × 97 orthogonal curvi- linear grid points in the horizontal, with variable spacing from 2 km near the head of the GoT to 55 km at the eastern boundary, with 10 sigma levels in the vertical conforming to realistic bottom topography. The model predicts the SSE up to 10.65 cm at Bangnara Estuary lo- cated in the south of Thailand. The sea level records of Buoy stations provided by the Marine Department and Hydrographic Department of Thailand are used as the observations in this study. The correlations between simulated SSE and observed SSE have strong positive corre- lations over the totally buoy stations of the GoT. Therefore the POM model is suitable to use to find the tendency of SSE in the GoT. From testing hypothesis, the research result meets that at the same position and same time the simulated SSE had a difference from the observed sea level not exceed RMSE at 0.05 significant levels.

Mathematics Subject Classification: 62H20, 65C20, 68U20

Keywords: Gulf of Thailand, buoy station, sea surface elevation, numer- ical ocean modeling, typhoon Muifa 752 N. Aschariyaphotha et al.

1 Introduction

The Gulf of Thailand (GoT) is located in Southeast Asia immediately to the west of the (SCS). The Gulf is a semi-enclosed sea that mea- sures approximately 400-km by 800-km, covering an area of about 320,000 square kilometers. Its location in the global map is between 6◦Nto14◦N lati- tude and 99◦E to 105◦E longitude surrounded by the Kingdom of Cambodia, Malaysia, the Kingdom of Thailand and the Socialist Republic of (Figure 1). It is a part of the Sunda Shelf, which is a submerged connection between Southeast Asia, Malaysia, Sumatra, Java, and Borneo, and is rela- tively shallow. The mean and maximum depths in the central part in the GoT are about 45 m and 80 m, respectively. This makes water exchange slow, and the strong water inflow from the rivers make the Gulf low in salinity and rich in sediments. The living and non-living resources of the Gulf are great value to the people of the four littoral countries. The management of fisheries, oil and gas resource development is in progress in this region. The numerical ocean modeling technologies have been developed in many countries. Blumberg and Mellor (1987) developed a three dimensional coastal ocean circulation model with a free surface, time-dependent, and which was named Princeton Ocean Model (POM). Now POM is one of the widely used ocean models. The POM describes in the rectangular coordinate system and uses the splitting method to isolate the barotropic gravity wave terms from baroclinic terms. There are several POM-based studied in the SCS connected to the GoT, e.g. Chu et al. (1998) used POM to simulate the circulation and thermohaline variability for the SCS covering the GoT. They verified that the wind effect is the key factor for the generation of the SCS deep basin warm/cool eddy and that lateral boundary forcing is the major factor for the formation of the strong western boundary currents. Yang et al. (2002) used the POM to study the seasonal mean SCS circulation and its formation mechanisms. It reproduced well the observed sea surface height (SSH) annual cycle, and the sensitivity experiments show that the wind forcing dominates the seasonal variability of the SCS SSH, while the buoyancy forcing is of minor importance. Aschariyaphotha et al. (2008) used the POM to investigate the effects of wind force and open boundary conditions in the GoT. The results show that wind force is the important factor for generating the circulation in the GoT. The lateral velocity at the open boundaries is of considerable importance to the current circulation for the rainy season and the end of the rainy season, with insignificant effect for the winter and summer seasons. The GoT is subjected to the monsoon system of the western North Pacific Ocean or the SCS. Since the GoT is shallow and locate mainly within land, it takes the effect of tropical storms. Thus the typhoon period time should be Statistical analysis of SSE for the GoT during typhoon Muifa 753 used to illustrate the model in the simulation phase. Typhoon Muifa (Figure 2) occurring in the GoT during November 21-28, 2004 is selected to be the case study of the model prediction. The model was run with the spin-up results and real-time wind. Typhoon Muifa first became a tropical depression on November 11, 2004, south of the Caroline Islands in the West Pacific. The storm moved steadily northwest, passing well north of Palau and just south of the Yap Islands before entering the Philippine Sea. Until November 25, 2004, typhoon Muifa slowly organized into a tropical storm and entered Suratthani province with a max- imum wind speed of 55 km/hr. Very few researches have been made about typhoon Muifa and its storm surge in the GoT. Cheng, K. F. (2006) used Wavewatch-III (WW3) to study the response of the SCS to typhoon Muifa. The sea level records of buoy stations are provided by the Marine Depart- ment (MD) and the Hydrographic Department of Thailand. There are 6 buoy stations from the MD and 2 buoy stations from the Hydrographic Department. The objective of this paper is to simulate the sea surface elevation (SSE) during the attack of typhoon Muifa for the GoT at each buoy station by using the POM model. Comparison of tendency between the SSE from the model and the observations is investigated. The paper also aims to test the hypothesis of the error of the model for simulating the SSE.

2 Model Description

POM model is the time-dependent, primitive equation model on a three- dimensional grid in Cartesian coordinates and vertical sigma coordinate, which have all been modified to simulate the current circulation and sea surface ele- vation in the GoT. The model includes hydrostatic and Boussinesq approxima- tions. The equations of the three dimensional primitive model are as follows: du 1 ∂p ∂ ∂u = − + fv + Amv + Fx (1) dt ρ0 ∂x ∂z ∂z dv 1 ∂p ∂ ∂v = − − fu+ Amv + Fy (2) dt ρ0 ∂y ∂z ∂z ∂w ∂u ∂v + + = 0 (3) ∂z ∂x ∂y ∂p = −ρg (4) ∂z dT ∂ ∂T = Ahv + FT (5) dt ∂z ∂z dS ∂ ∂S = Ahv + FS (6) dt ∂z ∂z 754 N. Aschariyaphotha et al.

ρ = ρ(T, S, z), (7) where x, y and z represent the eastward, northward and upward directions, respectively; u, v and w are the corresponding current components in the east- ward, northward and upward directions, respectively; t is time; ρ0 is the ref- erence density of the sea water; p is the pressure; ρ is the density of sea water given by UNESCO (1981); T is the potential temperature; S is the salinity; f = 2Ω sin θ is the Coriolis parameter; Ω is the angular speed of the earth’s rotation (7.2921 × 10−5 rad s−1); θ is the latitude; g is the magnitude of grav- itational acceleration; Amv is the coefficient of vertical eddy viscosity, and Ahv is the coefficient of vertical eddy diffusivity. The vertical mixing coefficients, Amv and Ahv, are obtained by appealing to a second order turbulence closure scheme, as described by Mellor and Yamada (1982). The governing equations are then transformed from z-coordinate (x,y, z,t) to the vertical sigma coordinate (x∗,y∗,σ,t∗) by the relation, z − η x∗ = x, y∗ = y, σ = and t∗ = t, (8) H + η where H(x, y) is the bottom topography, and η is the sea surface elevation. The value of σ ranges from σ =0atz = η to σ = −1atz = −H(x, y). The mode splitting method (Simons, 1974; Madala and Piacsek, 1977) is used to separate out barotropic mode from baroclinic mode, but the model did not include tidal forcing and river outflow. For the model grid, the rectangu- lar horizontal grid was replaced by an orthogonal curvilinear grid to improve coastline representations with 37 × 97 horizontal grid points. The horizontal grid spacing has ranged from 2 km to 55 km. The vertical sigma coordinate has 10 levels. Aschariyaphotha et al. (2004) has described the techniques that generate the model grid and interpolated the initial data by using cubic spline and bilinear interpolations.

2.1 Initialization The bathymetry for the model was derived from the digital bathymetric data base 5-min (DBDB5) binary data file developed by the Naval Oceanographic Office (NAVOCEANO) with 1/12 degree longitude-latitude grided resolution. The numerical model integration was divided into spin up and simulation phases. In the spin up phase of the model run, the model was integrated with all three components of velocity initially set to zero, while the tem- perature and salinity were indicated by interpolating climatological monthly mean wind, restoring-type surface temperature and salinity and climatological monthly mean freshwater flux taken from the European Centre for Medium- Range Weather Forecasts (ECMWF). The drag coefficient for wind stress (CD) Statistical analysis of SSE for the GoT during typhoon Muifa 755

suggested by Matthias and Godfrey (1994) is used in this paper. The CD is calculated by CD = Min{0.001 + 0.00007|Va|, 0.0025}. (9)

For spin up phase of the first model run, the wind velocities, Va, were taken from ECMWF climatological monthly mean wind velocity measured from 10 m above the sea surface. On the bottom of the Gulf, the normal gradients of T and S are zero, therefore there are no heat fluxes and salt fluxes across this boundary. A coefficient of bottom friction, Cz, taken from Mellor (2004) is calculated by ⎧ ⎫ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎨ κ2 ⎬ Cz = Max  , 0.0025 , (10) ⎪ 2 ⎪ ⎪ (1 + σkb−1)H ⎪ ⎩⎪ ln ⎭⎪ z0 where κ is the Von Kraman constant, H is the bottom topography, and z0 is the bottom roughness. In this study, κ is taken to be 0.4, while z0 =0.01 m is used from Weatherly and Martin (1978). Side boundary conditions, which are open boundary conditions, at land masses are the most simple, including no-slip, non-thermal flow and zero fluxes of heat and salt. There exist two types of open boundaries, which are inflow and outflow. Whenever the inflow occurs at the boundary of the model domain. The boundary condition is applied as follows: ∂ ∂ (T,S)+u (T,S) = 0 (11) ∂t ∂x ∂ ∂ (T,S)+v (T,S) = 0 (12) ∂t ∂y ∂u ∂u ± ci = 0 (13) ∂t ∂x ∂v ∂v ± ci =0, (14) ∂t ∂y where ci is the baroclinic phase speed. While outflow occurs at the boundary of the domain, the radiation condi- tion obtained from Blumberg and Mellor (1987) is applied as follows:

∂Θ ∂Θ + Un = 0 (15) ∂t ∂n where Θ can be taken to be T, S, u or v, and the subscript n is the direction normal to the boundary. For the velocities normal to land, the lateral boundary conditions are set to zero. The velocities and sea surface elevations at the 756 N. Aschariyaphotha et al. open lateral boundary may be taken from observations if observed current and elevation data are available. In the case where there are no available observed data, the lateral boundary conditions will be obtained from the other ocean model’s results. The numerical model integration was divided into spin-up and simulation phases. In the spin-up phase of the model run, the model was integrated with all three components of velocity initially set to zero, and temperature and salinity were indicated by interpolating climatological monthly mean fields from Levitus94 (Levitus and Boyer, 1994; Levitus et al. 1994) to the model grid. In addition, wind stress was calculated from the European Centre for Medium-Range Weather Forecasts (ECMWF) climatological monthly mean wind. The lateral boundary conditions of sea surface elevation and velocity were prescribed from You’s model outputs (You et al., 2001, 2002; Li and You, 2003) and the radiation condition. It was run until all variables were stable, and then the results will be used as initial data for the simulations. After spin-up phase, the initial data for typhoon Muifa simulation was wind field measured from 10 m above sea surface provided by ECMWF. The 6-h wind fields with 1.125 × 1.1213 degree longitude-latitude grid resolution, and were interpolated by using cubic spline and bilinear interpolations to specify wind components at the model grid point.

2.2 Integration The computation starts at 0000UTC on November 21, 2004 and ends at 2300UTC on November 28, 2004. The wind forcing field was added every 6 h, and the results were saved every 1 h. Time steps used for typhoon Muifa event were the internal time step 300 s and 10 s for the external time step.

3 Buoy Stations

The sea level records of Buoy stations are provided by the MD and Hydro- graphic Department of Thailand. Six buoy stations from the MD are Prasae, Langsuan, Sichol, Pakpanang Estuary, Pattani Estuary and Bangnara Estu- ary. Two buoy stations from the Hydrographic Department are Huahin and Ko Prab. Table 1 shows the positions of buoy stations. The simulated SSE at buoy locations are examined to compare with the observations.

4 Correlation Analysis

The correlation analysis is used in this study to examine the possible relation of simulated SSE and observed SSE for each tide gauge station. Several statis- Statistical analysis of SSE for the GoT during typhoon Muifa 757

Table 1: The positions of buoy stations

Station name Station point Huahin 99◦E 57 48 12◦N 34 22 Ko Prab 99◦E 26 18 9◦N 15 47 Prasae 101◦E 42 21 12◦N 41 41 Langsuan 99◦E 09 38 9◦N 56 38 Sichol 99◦E 55 07 9◦N 00 45 Pakpanang Estuary 100◦E 12 08 8◦N 21 11 Pattani Estuary 101◦E 14 57 6◦N 54 08 Bangnara Estuary 101◦E 49 48 6◦N 25 26

tics can be employed to measure the correlation between two variables. The parameter r is denoted as the linear correlation coefficient [10]. The linear correlation coefficient is descriptive measure of the strength of the linear rela- tionship between two variables. The linear correlation coefficient, r,ofn data points can be computed from the formula n xy − x y r = . (16) n x2 − ( x)2 n y2 − ( y)2

The linear correlation coefficient, r, always lies between -1 and 1. Values of r close to -1 or 1 indicate a strong linear relationship between the variables and that the variable x is a good linear predictor of the variable y. On the other hand, values of r near 0 indicate a weak linear relationship between the variables and that the variable x is a poor linear predictor of the variable y. Positive values of r suggest that the variables are positive linearly correlated, meaning that y trends to increase linearly as x increases, with the tendency being greater the closer that r is to 1. Negative values of r suggest that the variables are negative linearly correlated, meaning that y trends to increase linearly as x increases, with the tendency being greater the closer that r is to -1. In this research we determine the variable x as daily mean simulated SSE, the variable y as daily mean observed SSE, and n is the total number of day during the period of typhoon Muifa.

5 Results and Discussion

The simulated SSE and observed SSE during November 21-28, 2004 at buoy stations are shown in Figure 3 and Figure 4, respectively. A peak of simulated SSE is 10.65 cm occurred at Bangnara Estuary on November 28, 2004. The 758 N. Aschariyaphotha et al. correlations between the simulated SSE and observed SSE are shown in Table 2. Relationship between the simulated SSE and observed SSE during period of typhoon Muifa has significant strong positive correlations for all buoy stations. Huahin, Pattani Estuary and Bangnara Estuary have positive correlation in the range of 0.5 to 0.75. Ko Prab, Prasae, Langsuan, Sichol and Pakpanang Estuary have strong positive correlation greater than 0.75.

Table 2: The correlations between the simulated SSE and observed SSE at buoy stations

Station name Correlation coefficient Huahin 0.6413 Ko Prab 0.8383 Prasae 0.7873 Langsuan 0.8622 Sichol 0.8068 Pakpanang Estuary 0.9265 Pattani Estuary 0.5729 Bangnara Estuary 0.5647

The results of the testing hypothesis, H0 : μ2 − μ1 ≤ RMSE, with t-test are shown in Table 3. μ2 is the mean of observed SSE; μ1 is the mean of simulated SSE, and RMSE is the Root Mean Square Error. The results show that the value predicts accepts the hypothesis of difference μ2 and μ1 which is not greater than the RMSE at 0.05 significant levels for all locations.

Table 3: The RMSE of the simulated SSE and testing hypothesis of μ2 − μ1 ≤ RMSE

Station name RMSE T p-value Huahin 27.65 -0.26 0.80 Ko Prab 46.47 0.13 0.90 Prasae 19.86 -2.10 0.07 Langsuan 25.25 -0.13 0.90 Sichol 6.75 -0.87 0.42 Pakpanang Estuary 23.58 -0.10 0.92 Pattani Estuary 26.35 -0.09 0.93 Bangnara Estuary 30.59 -0.07 0.95 Statistical analysis of SSE for the GoT during typhoon Muifa 759

6 Conclusion

The SSE with the rise of typhoon Muifa on November 21-28, 2004 are sim- ulated by POM model and the errors between the model simulation and ob- servation at eight buoy stations are examined. This research examines the correlation between simulated SSE and observed SSE from buoy stations. It can be concluded that the correlations between simulated SSE and observed SSE had strong positive correlations over the totally buoy stations of the GoT. This correlation associated an increasing (positive) trend of SSE in the GoT. Therefore the POM model is suitable to use to find the tendency of SSE in the GoT. The research result meets that at the same position and same time the simulated SSE had a difference from the observed sea level not exceed RMSE at 0.05 significant levels.

ACKNOWLEDGEMENTS. I expresses my sincere thanks to the De- partment of Mathematics, King Mongkut’s University of Technology Thonburi (KMUTT) for the facilities during this research. The buoy data sets were kindly provided by the MD and Hydrographic Department of Thailand.

References

[1] A.F. Blumberg and G.L. Mellor, A description of a three-dimensional coastal ocean circulation model, Three Dimensional Coastal Ocean Models, ed. by N.S. Heaps, American Geophysical Union, 1987.

[2] G. Weatherly and P.J. Martin, On the structure and dynamics of the ocean bottom boundary layer, Journal of Physical Oceanography, 8 (1978), 557 - 570.

[3] G.L. Mellor and T. Yamada, Develop of a turbulence closure model for geophysical fluid problems, Reviews of Geophysics and Space Physics, 20 (1982), 851 - 875.

[4] G.L. Mellor, Users Guide for a Three-Dimensional, Primitive Equation, Numerical Ocean Model, Princeton University, New Jersey, 2004.

[5] H.J. Yang, Q. Liu, Z. Liu, D. Wang and X. Liu, A general circulation model study of the upper ocean circulation of the South China Sea, Jour- nal of Geophysical Research, 107 (2002), C7.

[6] JTWC, Annual Tropical Cyclone Report, Available online: http://www.npmoc.navy.mil/jtwc/atcr/2004atcr/, 2004. 760 N. Aschariyaphotha et al.

[7] K.F. Cheng, Typhoon effects on the South China Sea wave characteristics during winter monsoon, Master Thesis, (2006), 115. [8] N. Aschariyaphotha, P. Wongwises and U.W. Humphries, Interpolation of climatological data to numerical ocean modeling for the Gulf of Thailand, Proceedings of the Joint International Conference on Sustainable Energy and Environment, (2004), 784 - 789. [9] N. Aschariyaphotha, P. Wongwises, S. Wongwises, U.W. Humphries and X.B. You, Simulation of seasonal circulations and thermohaline variabili- ties in the Gulf of Thailand, Advances in Atmospheric Sciences, 25 (2008), 489 - 506. [10] N.A. Weiss, Introductory Statistics, Addison Wesley Publishing Company, USA., 1995. [11] P.C. Chu, Y. Chen and S. Lu, Wind-driven South China Sea deep basin warm-core/cool-core eddies, Journal of Oceanography, 54 (1998), 347 - 360. [12] R.F. Li and X.B. You, The development and application of the oceanic general circulation model Part II: the regional oceanic general circulation model, Chinese Journal of Atmospheric Sciences, (2003), 729 - 739. [13] R.V. Madala and S.A. Piacsek, A semi-implicit numerical model for baro- clinic oceans, Journal of Computing Physics, 23 (1977), 167 - 178. [14] S. Levitus and T. Boyer, World Ocean Atlas, Vol 4: Temperature, NOAA Atlas NESDIS, U.S. Government Printing Office, Wasington, D.C., 1994. [15] S. Levitus, R. Burgett and T. Boyer, World Ocean Atlas, Vol 3: Salinity, NOAA Atlas NESDIS, U.S. Government Printing Office, Wasington, D.C., 1994. [16] T. Matthias and J.S. Godfrey, Regional Oceanography: An Introduction, Pergamon, Australia, 1994. [17] T.J. Simons, Verification of numerical models of Lake Ontario, Part I: Cir- culation in spring and early summer, Journal of Physical Oceanography, 4 (1974), 507 - 523. [18] UNESCO, Tenth Rep. of the joint panel on oceanographic tables and standards, UNESCO Tech. Pap. in Marine Science, 36 (1981). [19] X.B. You, R.F. Li, M. Zhang and Q.C. Zeng, Numerical calculation of the wintertime circulation in the South China Sea by using a baroclinic model, Acta Oceanologia Sinica, (2001), 1 - 10. Statistical analysis of SSE for the GoT during typhoon Muifa 761

[20] X.B. You, R.F. Li and Q.C. Zeng, Numerical simulation of the current system in the China Sea by using a nested mesh model, Proceedings of the Nineth Asian Congress of Fluid Mechanics, (2002).

Figure 1: Geography and bathymetry (m) of the Gulf of Thailand.

Received: October, 2012 762 N. Aschariyaphotha et al.

Figure 2: The best track passage of typhoon Muifa (2004) (JTWC, 2004). Statistical analysis of SSE for the GoT during typhoon Muifa 763

Simulated SSE (cm)

14 Huahin Lang Suan Ko Prab 12 Prasae

10

8

6

4

2

0

−2 21/11/04 22/11/04 23/11/04 24/11/04 25/11/04 26/11/04 27/11/04 28/11/04

Simulated SSE (cm)

Si Chol 14 Pakpanang Pattani 12 Bangnara

10

8

6

4

2

0 21/11/04 22/11/04 23/11/04 24/11/04 25/11/04 26/11/04 27/11/04 28/11/04

Figure 3: The simulated sea surface elevation (cm) during November 21-28, 2004 at buoy stations. 764 N. Aschariyaphotha et al.

Obseved SSE (cm) 80 Huahin Lang Suan 70 Ko Prab Prasae 60

50

40

30

20

10

0 21/11/04 22/11/04 23/11/04 24/11/04 25/11/04 26/11/04 27/11/04 28/11/04

Observed SSE (cm)

Si chol 50 Pakpanang Pattani Bangnara 40

30

20

10

0

21/11/04 22/11/04 23/11/04 24/11/04 25/11/04 26/11/04 27/11/04 28/11/04

Figure 4: The observed sea surface elevation (cm) during November 21-28, 2004 at buoy stations.